
From michawe@ifi.uio.no  Wed Jul  1 01:16:23 2009
Return-Path: <michawe@ifi.uio.no>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id B6F163A6810; Wed,  1 Jul 2009 01:16:23 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.599
X-Spam-Level: 
X-Spam-Status: No, score=-6.599 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id nP6igwppcBrx; Wed,  1 Jul 2009 01:16:22 -0700 (PDT)
Received: from mail-out2.uio.no (mail-out2.uio.no [129.240.10.58]) by core3.amsl.com (Postfix) with ESMTP id 514A13A6802; Wed,  1 Jul 2009 01:16:22 -0700 (PDT)
Received: from mail-mx5.uio.no ([129.240.10.46]) by mail-out2.uio.no with esmtp (Exim 4.69) (envelope-from <michawe@ifi.uio.no>) id 1MLuzi-0001DJ-Eg; Wed, 01 Jul 2009 10:16:42 +0200
Received: from div-8021x-dhcp050.uio.no ([193.157.176.59]) by mail-mx5.uio.no with esmtp  (Exim 4.69) (envelope-from <michawe@ifi.uio.no>) id 1MLuzh-00065m-VT; Wed, 01 Jul 2009 10:16:42 +0200
From: Michael Welzl <michawe@ifi.uio.no>
To: dccp@ietf.org, tsvwg@ietf.org
In-Reply-To: <030C089D-DDA7-422B-B0B8-4922CC3D6B1A@iki.fi>
References: <128F7A72-30C0-46DD-9B7F-6FBB2B038CFA@iki.fi> <2f7f60b7442cf1efb11f632267f056d0.squirrel@webmail.uio.no> <97D8957C5565BB41912C3F958914C49F0AFBD972@sonusmail06.sonusnet.com> <9eebfffbafd6ec369430d9019d59123a.squirrel@webmail.uio.no> <030C089D-DDA7-422B-B0B8-4922CC3D6B1A@iki.fi>
Content-Type: text/plain
Date: Wed, 01 Jul 2009 10:16:57 +0200
Message-Id: <1246436217.4131.34.camel@localhost.localdomain>
Mime-Version: 1.0
X-Mailer: Evolution 2.10.3 (2.10.3-10.fc7) 
Content-Transfer-Encoding: 7bit
X-UiO-Ratelimit-Test: rcpts/h 6 msgs/h 2 sum rcpts/h 16 sum msgs/h 4 total rcpts 1591 max rcpts/h 33 ratelimit 0
X-UiO-Spam-info: not spam, SpamAssassin (score=0.0, required=5.0, autolearn=disabled, none, uiobl=NO, uiouri=_URIID_)
X-UiO-Scanned: ED802925DB88B9B5348E44071AA071AB5183F4AE
X-UiO-SPAM-Test: remote_host: 193.157.176.59 spam_score: 0 maxlevel 80 minaction 2 bait 0 mail/h: 2 total 361 max/h 15 blacklist 0 greylist 0 ratelimit 0
Cc: dragana.damjanovic@uibk.ac.at
Subject: [tsvwg] MulTFRC (draft-welzl-multfrc-00.txt)
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 01 Jul 2009 08:16:23 -0000

Hi,

In a recent email exchange, I mentioned that we designed
a congestion control mechanism which we call "MulTFRC",
and which we'd like to ask this group to adopt.
Tom answered that this seems to be quite in line with
the group's charter.

Since the group does not meet in Stockholm, it was
suggested that we ask for a presentation slot in
tsvwg or tsvarea, and I just did (for tsvwg).


I wanted to give people on the lists an update:
when I first announced it, we only had a paper, not
a draft. Now we have both; we also have an ns-2
implementation and real-life code. It's all available
from this page:
http://www.welzl.at/research/projects/multfrc/index.html

The draft is:
draft-welzl-multfrc-00.txt
http://www.ietf.org/internet-drafts/draft-welzl-multfrc-00.txt

Discussion should take place in DCCP, I think.


Cheers,
Michael


On Tue, 2009-06-02 at 22:42 +0300, Pasi Sarolahti wrote:
> Hi,
> 
> On Jun 1, 2009, at 19:14, michawe@ifi.uio.no wrote:
> 
> >> Even though we won't be meeting in Stockholm, if you think you can  
> >> get a
> >> draft in time we can explore getting some time for you to present at
> >> tsvarea or tsvwg.  Pasi, do you agree?  Especially since you'll be in
> >> Stockholm and I won't :-).
> >
> > We could definitely have a draft in time - this would be
> > a draft of only our mechanism, not a CCID, following the
> > approach of the TFRC spec, because this is an easier
> > first step in any case. Turning this into a CCID description
> > in case people don't want a standalone MulTFRC draft
> > shouldn't be a problem.
> >
> > So, I would greatly appreciate getting some time in
> > tsvarea or tsvwg!
> 
> I think it would be fine to discuss this in one of the above groups,  
> but the chairs of the groups would need to be consulted for a  
> presentation slot.
> 
> Are there plans to have an iccrg session in Stockholm? That wouldn't  
> be bad place, either, to present MulTFRC for the first time. Or if  
> none of the above works, I think it would also be possible to arrange  
> a kind of unofficial Bar-DCCP meeting to discuss this and possible  
> other new topics, even though we didn't ask to schedule an official  
> DCCP meeting this time.
> 
> In any case, I'd also like to encourage preparing a draft by the  
> Stockholm meeting.
> 
> - Pasi
> 


From root@core3.amsl.com  Wed Jul  1 23:30:01 2009
Return-Path: <root@core3.amsl.com>
X-Original-To: tsvwg@ietf.org
Delivered-To: tsvwg@core3.amsl.com
Received: by core3.amsl.com (Postfix, from userid 0) id 6B6D13A6D15; Wed,  1 Jul 2009 23:30:01 -0700 (PDT)
From: Internet-Drafts@ietf.org
To: i-d-announce@ietf.org
Content-Type: Multipart/Mixed; Boundary="NextPart"
Mime-Version: 1.0
Message-Id: <20090702063001.6B6D13A6D15@core3.amsl.com>
Date: Wed,  1 Jul 2009 23:30:01 -0700 (PDT)
Cc: tsvwg@ietf.org
Subject: [tsvwg] I-D Action:draft-ietf-tsvwg-port-randomization-04.txt
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 02 Jul 2009 06:30:01 -0000

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Transport Area Working Group Working Group of the IETF.


	Title           : Port Randomization
	Author(s)       : M. Larsen, F. Gont
	Filename        : draft-ietf-tsvwg-port-randomization-04.txt
	Pages           : 31
	Date            : 2009-07-01

Recently, awareness has been raised about a number of "blind" attacks
that can be performed against the Transmission Control Protocol (TCP)
and similar protocols.  The consequences of these attacks range from
throughput-reduction to broken connections or data corruption.  These
attacks rely on the attacker's ability to guess or know the five-
tuple (Protocol, Source Address, Destination Address, Source Port,
Destination Port) that identifies the transport protocol instance to
be attacked.  This document describes a number of simple and
efficient methods for the selection of the client port number, such
that the possibility of an attacker guessing the exact value is
reduced.  While this is not a replacement for cryptographic methods
for protecting the connection, the described port number obfuscation
algorithms provide improved security/obfuscation with very little
effort and without any key management overhead.  The algorithms
described in this document are local policies that may be
incrementally deployed, and that do not violate the specifications of
any of the transport protocols that may benefit from them, such as
TCP, UDP, UDP-lite, SCTP, DCCP, and RTP.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-tsvwg-port-randomization-04.txt

Internet-Drafts are also available by anonymous FTP at:
ftp://ftp.ietf.org/internet-drafts/

Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Message/External-body;
	name="draft-ietf-tsvwg-port-randomization-04.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID: <2009-07-01231803.I-D@ietf.org>


--NextPart--

From root@core3.amsl.com  Sun Jul  5 13:15:01 2009
Return-Path: <root@core3.amsl.com>
X-Original-To: tsvwg@ietf.org
Delivered-To: tsvwg@core3.amsl.com
Received: by core3.amsl.com (Postfix, from userid 0) id 95EAF3A6BFE; Sun,  5 Jul 2009 13:15:01 -0700 (PDT)
From: Internet-Drafts@ietf.org
To: i-d-announce@ietf.org
Content-Type: Multipart/Mixed; Boundary="NextPart"
Mime-Version: 1.0
Message-Id: <20090705201501.95EAF3A6BFE@core3.amsl.com>
Date: Sun,  5 Jul 2009 13:15:01 -0700 (PDT)
Cc: tsvwg@ietf.org
Subject: [tsvwg] I-D Action:draft-ietf-tsvwg-sctp-strrst-00.txt
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 05 Jul 2009 20:15:01 -0000

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Transport Area Working Group Working Group of the IETF.


	Title           : Stream Control Transmission Protocol (SCTP) Stream Reconfiguration
	Author(s)       : R. Stewart, et al.
	Filename        : draft-ietf-tsvwg-sctp-strrst-00.txt
	Pages           : 20
	Date            : 2009-07-05

Many applications that desire to use SCTP have requested the ability
to "reset" a stream.  The intention of resetting a stream is to start
the numbering sequence of the stream back at 'zero' with a
corresponding notification to the upper layer that this act as been
performed.  The applications that have requested this feature
normally desire it so that they can "re-use" streams for different
purposes but still utilize the stream sequence number for the
application to track the message flows.  Thus, without this feature,
a new use on an old stream would result in message numbers larger
than expected without a protocol mechanism to "start the streams back
at zero".  This documents presents also a method for resetting the
transport sequence numbers and all stream sequence numbers.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-tsvwg-sctp-strrst-00.txt

Internet-Drafts are also available by anonymous FTP at:
ftp://ftp.ietf.org/internet-drafts/

Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Message/External-body;
	name="draft-ietf-tsvwg-sctp-strrst-00.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID: <2009-07-05130636.I-D@ietf.org>


--NextPart--

From gorry@erg.abdn.ac.uk  Tue Jul  7 00:08:19 2009
Return-Path: <gorry@erg.abdn.ac.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 0092228C227 for <tsvwg@core3.amsl.com>; Tue,  7 Jul 2009 00:08:19 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.556
X-Spam-Level: 
X-Spam-Status: No, score=-2.556 tagged_above=-999 required=5 tests=[AWL=0.043,  BAYES_00=-2.599]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id tfMSaC6j1+HG for <tsvwg@core3.amsl.com>; Tue,  7 Jul 2009 00:08:16 -0700 (PDT)
Received: from erg.abdn.ac.uk (dee.erg.abdn.ac.uk [IPv6:2001:630:241:204:203:baff:fe9a:8c9b]) by core3.amsl.com (Postfix) with ESMTP id F2BDA28C1BA for <tsvwg@ietf.org>; Tue,  7 Jul 2009 00:08:15 -0700 (PDT)
Received: from Gorry-Fairhursts-Laptop-6.local (ra-gorry.erg.abdn.ac.uk [139.133.204.42]) (authenticated bits=0) by erg.abdn.ac.uk (8.13.4/8.13.4) with ESMTP id n6778YiZ017651 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES128-SHA bits=128 verify=NOT); Tue, 7 Jul 2009 08:08:34 +0100 (BST)
Message-ID: <4A52F472.3040500@erg.abdn.ac.uk>
Date: Tue, 07 Jul 2009 08:08:34 +0100
From: Gorry Fairhurst <gorry@erg.abdn.ac.uk>
Organization: The University of Aberdeen is a charity registered in Scotland,  No SC013683. 
User-Agent: Thunderbird 2.0.0.21 (Macintosh/20090302)
MIME-Version: 1.0
To: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
References: <4A43D532.6040906@erg.abdn.ac.uk> <200906261328.n5QDSuiM007908@bagheera.jungle.bt.co.uk>
In-Reply-To: <200906261328.n5QDSuiM007908@bagheera.jungle.bt.co.uk>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
X-ERG-MailScanner: Found to be clean
X-ERG-MailScanner-From: gorry@erg.abdn.ac.uk
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] Some comments on draft-ietf-tsvwg-ecn-tunnel-02]]
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
Reply-To: gorry@erg.abdn.ac.uk
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 07 Jul 2009 07:08:19 -0000

See a few responses to your comments in-line,

Gorry

Bob Briscoe wrote:
> Gorry,
> 
> Tx for such a thoro review. It seems nearly all the comments are about 
> poor description, rather than any protocol disagreements. Is this fair?
> 
Yes.

> The only substantive disagreement seems to be over whether ECT(1) 
> encapsulating  CE should raise an alarm at the egress.
> 
I am assuming that you think the role/use of tunnel negotiation is also 
a document organisation issue?

> Inline...
> 
> At 20:51 25/06/2009, Gorry Fairhurst wrote:
>> Bob,
>>
>> I've just read draft-ietf-tsvwg-ecn-tunnel-02, and have some comments.
>> Thanks for preparing this new version, which is certainly easier to read
>> (and seems a little shorter) than the previous one, I think as I said 
>> at the IETF, the general approach in this draft seems good to me.
>>
>> I've separated my comments into two parts. The first parts are comments
>> which I'd hope you may be able to respond to. The second parts I think 
>> may be editorial comments (but do ask questions, if you do not 
>> agree/understand).
>>
>> Best wishes,
>>
>> Gorry
>>
>>
>> ---
>> 1) Updates RFC4301?
>>
>> I do not understand whether the document updates RFC 4301, or is an
>> update that aligns with the changes made in RFC 4301. At some points,
>> the document says the former, whereas at other points I interpret it as
>> the latter. This needs to be very clear in the abstract and the document
>> text.
> 
> I thought I'd made this clear, but obviously not. I included the section 
> below which is a hint to the RFC editor to make this clear in the 
> headers and in the RFC index.
 >
Yes, I saw this, and this triggered my comment.

> /\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\ 
> 
> Additional changes to the RFC Index (to be removed by the RFC Editor):
> In the RFC index, RFC3168 should be identified as an update to RFC2003. 
> RFC4301 should be identified as an update to RFC3168.
> This specification updates RFC3168 and RFC4301.
> /\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\ 
> 
> 
> Updating 4301 isn't as bad as it sounds tho:
> * The tunnel ingress behaviour is in line with 4301.
> * The tunnel egress behaviour adds to 4301, but doesn't change any 4301 
> behaviours that currently exist. So compliance with this spec has no 
> backward compat issues.
> Therefore compliance can be called for only as it is needed (e.g. for PCN).
> 
OK, I was NOT raising an issue with updating RFC 4301.

(I expect the security area will say if they do have issues.)

In this case, I would have expected a section saying how RFC 4301 was 
updated, butdid not find a section.

>> ---
>> 2) Abstract
>>
>> I'd be happier if the abstract could call-out that this updates the 
>> behaviour of tunnel end points and middleboxes. That is say something 
>> so that people using and configuring these devices know to read this 
>> document.
> 
> I can say it updates the behaviour of *some* types of tunnel end-point. 
> But it would be too complicated for an abstract to say exactly which ones.
> 
Let's try something along these lines.

> 
>> I'd also like the abstract to be clear that there is more than one mode
>> and say something of this topic - I'm not sure what, because I still
>> have questions in this space (see below).
> 
> The mode stuff is not needed in most cases, and the actions in this RFC 
> make modes less likely to be needed. 
>
That is what I hoped.

> It's only for some cases that are 
> trying to be completely generic and dynamic. And even then, it's only 
> carrying through detailed stuff that was already in RFC3168. As it's not 
> the new thing, I'd say it's not appropriate for the abstract.
> 
OK. I'll see what the new abstract looks like.

>> The abstract should say the set of RFCs that are updated by the 
>> document (but not cite these as references), e.g.  This specification 
>> updates RFC3168 {and RFC4301 --- if it does}.
> 
> I'd rather get the RFC Editor to put "updates RFC3168 and RFC4301" in 
> the status header, than crowd the abstract.  (that's why I added the 
> note for the RFC Editor in S.6 "Changes from Earlier RFCs")
> 
Not sure I agree.

We can revisit this when get to WGLC.

>> ---
>> 3) Current status of PCN work
>>
>> Section 3 includes:
>> /  A  number of wasteful or convoluted work-rounds to this problem are
>>       being considered for standardisation by the PCN working group (see
>>       Appendix D), but by far the simplest approach is just to remove
>>       the covert channel blockages from tunnelling behaviour, that are
>>       now deemed unnecessary anyway./
>> - Is this still the case? If this document removes this constraint, can
>> we just say so?
> 
> It's still the case, because the PCN work can't assume this spec will 
> get through yet - last calls can raise probs even tho I've been careful 
> to ensure there are no back-compat issues for the security area.
> 
So, we will need to sort this out before we reach WGLC.

>> ---
>> 4) Recommended default and tunnel negotiation method for non-IPsec case?
>>
>> - The following set of points are all very much related, and I would 
>> like some more detail on what you envisage:
>> ---
>> Section 4.1
>> - The current section seems to step-back from actually recommending what
>> the mode that SHOULD be used when there is no tunnel negotiation. This
>> leaves me with some questions about  what a tunnel endpoint needs to do
>> by default.
>> - What do you think the document should say?
> 
> OK, I thought I'd made it clear, but obviously not.
>
> 4.1 refers forward to 5. on this issue (so as not to crowd the main 
> point with details that aren't usually necessary).
>
> 5.1 says "...the new normal ingress behaviour introduces no backward 
> compatibility isses with IKEv2 [RFC4306] IPsec [RFC4301] tunnels, and no 
> need for any new modes, options or configuration. "
> 
> I think I see the problem. I tend to start by describing the logic for 
> all cases, then say what this implies for each. Instead I should assert 
> what should be done in each case, then say why. I'll do the latter in 
> the next rev.
>
> It's a good point. Implementers just want to find the bits they need to 
> read and ignore the rest.
> 
OK, I think it would be good to see something that says what to do
in "normal" IP in IP tunnels, then be specific for IPsec, etc.

>> ---
>> Section 5:
>>   /Before switching to normal mode, a compliant tunnel ingress that does
>>    not know the egress ECN capability MUST negotiate with the tunnel
>>    egress./
>> -It is not clear to me *HOW* this tunnel negotiation would normally be
>> performed. Can you clarify?
> 
> Again, I've obviously made it hard-reading by giving the general logic 
> for all cases, rather than listing each common case. How about text 
> something like this to start off with:
> 
> "Both endpoints of an RFC4301 tunnel can be sure that the other end is 
> 4301-compatible, because the tunnel is only formed after key management 
> has been completed, at which point both ends will be RFC4301-compliant 
> by definition. Therefore, an IPsec tunnel ingress compliant with RFC4301 
> does not need to implement compatibility mode for legacy ECN tunnels.
> 
> Similarly, the implementer

- Is this the "implementor" of this spec. or is when tunnels are
configured in networks?

> of any (non-RFC4301) tunnel ingress that is 
> designed to be deployed at the same time as its corresponding tunnel 
> egress, will not need to implement compatibility mode for legacy ECN 
> tunnels.
> 
Again, I am not sure who is regarded as the implementor.

> However, an implementer might want to create a general-purpose tunnel 
> ingress that includes some form of tunnel endpoint discovery to 
> dynamically create and destroy tunnels to different tunnel egresses, 
> driven by the varying needs of arriving packets. If the tunnel discovery 
> process is arranged to only ever find tunnel egresses compliant with 
> normal mode of RFC3168 or a later ECN tunnelling specification, then the 
> ingress would not need to implement compatibility mode. But if earlier 
> tunnel egresses might be discovered, the ingress would need to implement 
> compatibility mode as well as normal mode, and also implement 
> appropriate negotiation as defined in RFC3168."
>
So, I still don't understand.

If we are saying how to *USE* this new RFC to be, is it possible to say
all tunnel discovery protocols SHOULD do something specific?

> Would this text be OK?
> 
>> I found this paragraph in Section 5.3
>> /   Therefore, a compliant tunnel ingress MUST at least implement `normal
>>    mode' and, if it might be used with arbitrary tunnel egress nodes, it
>>    MUST also implement `compatibility mode'./
> 
> Is this text a problem? Or are you saying it solves the above issues? Or 
> you want it said earlier?
> 
Not a problem, I thought this was clearer, and wondered whether this was
the crucial point.

>> ---
>> Section 5.3, para 2:
>> Section 5.3, para 4:
>> - Para 2 and later talks of safeguards, and if I understand it makes a
>> requirement that tunnel ingress MUST start in compatibility mode if it
>> can not negotiate.
> 
> It doesn't say "if it cannot negotiate". It says if it "does not know 
> the egress ECN capability ".
> 
How may the tunnel "know" this?

>> How do tunnels find out this information? (see earlier).
> 
> It's easy if both ends of the tunnel are statically config'd (ie the 
> info is part of the config, and it's up to the sysadmin to get it right).
> 
OK, I understand that case.

> I'm trying to get rid of modes, but I have to allow for the same problem 
> RFC3168 had to allow for. So I have to allow the same negotiation as 
> RFC3168. I don't know if anyone does this, or how exactly they do it. 
> I've been told negotiation is frowned on. But I'm not adding anything 
> beyond what RFC3168 said.
> 
> I'll say "...negotiation as in RFC3168" if that helps.
> 
It's better.

> In S.6 I have clarified that RFC4301 IPsec implies the negotiation 
> machinery of RFC3168 isn't needed for an IPsec tunnel any more:
> 
> "The optional ECN Tunnel field in the IPsec security association 
> database (SAD) and the optional ECN Tunnel Security Association 
> Attribute defined in RFC3168 are no longer needed. The security 
> association (SA) has no policy on ECN usage, because all RFC4301 tunnels 
> now support ECN without any policy choice. "
 >
- That's clear.

>> - I am not sure how the tunnels we be configured outside of IPsec, so
>> could you help me understand how this negotiation will take place?

I think I still do not understand this.

>> ----------
>> 5) Informative status of section 4.3
>>
>> Section 4.3 is stated as informative
>> - Can you remind me why this section is informative?
>> - It seems that there  are strong messages here on the use of the
>> method. /should/ and /should  not/ are used many times in a
>> non-normative way. I'd like to explore whether the WG would like to make
>> these guidelines normative.
> 
> Maybe I don't understand customary procedure, but I imagine that 
> guidelines for future protocol design are generally informative, and 
> normative is only needed when a standards track doc is written that uses 
> the guidelines. As this doc is intended for standards track, I wanted to 
> make the boundary clear, where it had shifted into "informational land".
> 
> I don't believe guidelines for future protocol design mean anything more 
> than "you'll need to give good reasons, if you break these". Perhaps my 
> philosophy for encouraging evolvability doesn't match the mentality of 
> the typical IETFer these days tho, so discussion is welcome.
> 
I think it's fine to allow for evolution. It may also be OK to provide 
standards-track recommendations, where this impacts the interoperability.

>> ---
>> 6) Relation of section 6 with Updating IPsec
>>
>> Section 6:
>> - I do not understand this section. I'd have expected a specific section
>> on how this document updates IPsec - if it does, but I was not sure
>> whether this section was intended to be the update to IPsec, or whether
>> the changes relate to all tunnel end points (if so, how
> 
> This section is meant to summarise how the changes introduced in prior 
> sections update previous RFCs.
> 
> But the section is not organised by previous RFCs. It's organised by 
> function:
> - ingress
> - egress
> - negotation between the two (if present).
> 
> For each function, it says which aspects of previous RFCs are updated.
> 
> This was designed to help implementers, who would be working on each 
> function, having to pull together the relevant RFCs for that function.
> 

>> - The discussion here seems to assume IPsec and IKEv2, but I do not
>> understand the context, either more is needed or much less is needed.
> 
> It doesn't assume IPSec. It just says what has changed if you're doing 
> IPSec.
> 
> IKEv2 key negotiation is integral to RFC4301. But David Black told me 
> there could be corner cases that use manual keying instead.
> 
I'd agree.

> I could remove the stuff on the manual keying corner case. I'd like to 
> take David Black's advice on that. He implied if I put it in it will get 
> unnecessarily complicated given no-one does this in practice, altho 
> RFC4301 allows them to. But if I leave it out, someone could point out 
> that we haven't thought about it.
> 
> Does this help?
> 
I understand what you say.

>> ---
>> 7) Alarm raised when ECT(1) encapsulates CE.
>>
>> Section 6, bullet 3 / Figure 4:
>> - This says an alarm is generated, Is this really a problem?
> 
> In practice, implementers tend to drop unexpected combinations, on the 
> spurious pretext that there might be a security risk.
> 
> The idea behind recommending alarms is to give implementers a legitimate 
> way to not drop combinations just because are unexpected today, given 
> they might be used in the future. While still allowing operators to 
> monitor for possible attacks.
> 
OK

> Then, if there's a sudden rise in alarms of this type, the operator can 
> engage brain and say either
> - "Ah, that's because RFC7024 is starting to be implemented, we can turn 
> off those alarms" or
> - "Hmmm, there's not even an experimental standard that would cause this 
> strange combination, could this be a novel attack?"
> 
>> - I understand that ECT(0) is allowed in some cases to be re-mapped to
>> ECT(1).
> 
> If you know of these cases, then fine, I'll remove the need for an 
> alarm. But I wasn't aware of any. Any specifics?
> 
My thoughts were that there may be multi-level ECN implementations that 
remark in this way. My point was that if I am setting-up a tunnel, I 
would prefer not to be concerned about the usage of the codepoints.

>> This then makes me wonder why a packet that was remarked from
>> ECT(0) to ECT(1) then raises an alarm when the inner packet was CE
>> marked. Have I understood this correctly?
>> /outer ECT(1) with inner CE/
>> - To me, this also seems compatible with RFC 4301.
> 
> RFC4301 & RFC3168 define the same output (CE) as I do for this 
> combination. The issue is over whether the combination of inputs would 
> ever be *expected* to arise, given none of the other ECN RFCs could lead 
> to this happening.
> 
> I'm not aware of a valid ECN protocol transition that would produce this 
> combination, so I've suggested it should raise an alarm. RFC3168 had 
> some cases that raised alarms, but I just think it missed this one. If 
> I'm wrong, I'll happily retract.
> 
OK, let me check: Are there any PCN or other ECN variants (Such as 
multi-level ECN (as in RFC 4774) that can re-mark ECT(0) to ECT(1), 
under any conditions?

>> ---
>> 8) "Danger" of not forwarding congestion indications?
>>
>> Section 8:
>> /  In fact, ensuring congestion markings are not
>>    lost is itself another aspect of security, because if we allowed
>>    congestion notification to be lost, any attempt to enforce a response
>>    to congestion would be much harder./
>> - I accept that loss of congestion marking can impede congestion
>> control. I do no agree that this is accepted as a factual security
>> threat to the Internet.
> 
> Flooding attacks classify as a security threat (not information 
> security, but still security). Enforcing congestion control is a valid 
> approach for handling flooding attacks. If we allow explicit congestion 
> information to be thrown away, it makes both responding to it and 
> enforcing a response to it harder. That's all I'm saying. Surely that's 
> non-controversial?
> 
This warning is not at all controversial - it seems a quite reasonable 
thing to say, please could you just say more than the word "dangerous".

> I appreciate that a large part of the security community focuses on 
> using information security rather than congestion control enforcement to 
> solve flooding, but that doesn't mean they are right. And others 
> disagree with them.
> 
I personally agree with your statement that this is important to protect
the network service, and expect others in the TSV area to agree with this.

> 
>> ---
>> 9) Judgement on ECN-N?
>>
>> Section 8:
>> This last statement on ECN-N adoption is not required or neccessary for
>> this document. I would rather not see statements such as /and there has
>> been no evidence that anyone has implemented it beyond the author's
>> prototype./ introduced as a side item:
>>
>>    4.  The ECN nonce [RFC3540] currently has experimental status and
>>        there has been no evidence that anyone has implemented it beyond
>>        the author's prototype.
> 
> An RFC changes as it moves from draft to RFC. To start with it has to 
> justify what it's doing. As people accept the arguments, a lot of 
> justification text becomes removeable. I'd be happy to remove 
> justifications eventually (as text like this can get dated), but I'd 
> rather not until the arguments have got past the security area.
> 
> We have two conflicting choices to reconcile. This is one of the reasons 
> for the choice made. If we don't mention it, it weakens the case.
> 
If the working group really needs to make this statement in a working 
group draft, we need to raise this as a specific action.  I personally 
don't see the need for it here, and do not believe the update relies 
upon this.

> 
>> ---
>> 10) Size of Appendices.
>>
>> My review ran out of steam as I read the pages of appendices and I am
>> not currently persauded that most of the text in the appendices is
>> needed to publish the updated specification.
> 
> I'm also running out of steam here, as I'm meant to be going on holiday 
> for a week now. As my wife is politely suggesting a cup of water might 
> end up being spilled over my keyboard, I'll also be brief...
> 
> 
>> Section A.2:
>>
>> There are many places where the English could be improved. Examples from
>> the final para include:
>> /ingress confuses a proposed.../
>> /It ends up removing excessive amounts of traffic uncessarilly.../
>>
>> Section B seems to be related to the rationale for PCN, rather than the
>> use of ECN over tunnels. Is this really required in this specific 
>> document?
>>
>> B.1 seems to be a background to ECN
>> B.2 again reads like a paper, but the language is rather imprecise
>>
>> Appendix C seems OK, but is it needed?
>>
>> Appendix D seems to justify the work, but since this is now adopted, is
>> the text neccessary?
>> Section D.1 could be removed to improve the document.
>>
>> The appendices currently seem to read more like a technical paper than
>> an IETF PS. If these appendices really are required, then some editorial
>> work is required to make this a standards quality contribution, and
>> clarify what actually needs to be said. After this, someone will need to
>> review the statements about the discussed IETF protocols carefully -
>> perhaps by referral to the appropriate WGs.
> 
> There's a balance needed between recording why things were done and not 
> making the thing flowery and unweildy. Point taken. I'll give this some 
> attention in the next draft.
OK
> 
>> ---
>> 11) Relevance to GRE, IP-in-IP, L2VPN, etc, etc.
>>
>> - If this change is intended to be appropriate to a range of tunnel
>> behaviours, how do we make sure implementors clearly understand this?
>> - Perhaps it would be helpful to list a set of example encapsulations at
>> the front of the introduction in para 1.
> 
> In the rev between -01 & -02 I took out all the stuff on non-IP-in-IP 
> tunnels. It will probably re-appear in some future draft, but not this one.
> 
OK, that seems like a fair plan.

> If I've left any text in here that implies this is relevant beyond 
> IP-in-IP, pls point me at it. I thought I had taken it all out.
> 

> Tx for the nits below. I'll handle them.
> 
> 
> Bob
> 
> 
>> ----------------------------------------------------------------------
>> * NiTs
>> ---
>> Consider changing all:
>> /doesn't/
>> - to say /does not/
>> ---
>> Abstract:
>> /in order that these combinations may be usefully/
>> - could we say /to allow these combination to be usefully/
>> ---
>> Introduction, para 1:
>> /Three main change/
>>                  ^
>>  - please insert 's'
>> ---
>> para after table 1,
>> /it must follow rules to ensure marking is propagated.../
>>    ^^^^^
>>  - could say 'needs' to avoid conflict with RFC2119 keywords?
>> ---
>> /The newly specified behaviours/
>>  ^^^^^^^^^^^^^^^^^^^
>> - Please say /The new behaviours specified in this document.../
>> ---
>> /   The rules for constructing the ECN field to be forwarded after tunnel
>>    decapsulation ensure this happens, but they are not wholly/
>> - That's the rules in the documents that this document updates?
>> - Some references would be helpful here.
>> ---
>> Para starting /as a result,/
>> The following text, while true, will read oddly once published. I think
>> this sentence could be usefully rephrased to be more objective:
>> /As a result, an unfortunate sequence of standards actions has left us
>>    with nearly the worst of all possible combinations of outcomes,
>>    despite the best endeavours of everyone concerned./
>> - later in the para /still/ will likely be changed by publishing this
>> document and /now is/ was true at the time it was written, but not for
>> evermore.
>> ---
>> 1.1 Scope, para 2:
>> - should this refer to RFC 2983 for example?
>> ---
>> 1.2 bullet 1:
>> /current tunneling behaviours/
>> - legacy methods? - It would be clearer to explicitly reference the
>> specs. here.
>> ---
>> section 3, para 1:
>> /It merely recaps/
>>    ^^^^^^^ - could we delete this word?
>> ---
>> Figure 1
>> - final line has a spurious 'e' character.
>> ---
>> 3 paras after figure 1:
>> /But.../ is used twice at the start of a sentence. Consider deleting the
>> firs case, and replacing the second with /However,/
>> ---
>> Page 11, and following:
>> - The covert channel appears several times. I think it is good to
>> document this and explain the implications, but it seems to be repeated
>> several times. It also appears again at the end of section 9.
>> ---
>> End of section 3, first bullet, para 1:
>> /Firstly/
>> - Consider using /First,/
>> ---
>> Section 3.1, end:
>> / To ensure copying rather than resetting CE on ingress will not cause
>>    unintended side-effects, Appendix A assesses whether either harm any
>>    security, control or management functions. /
>> /It finds that resetting
>>    CE makes life difficult in a number of directions/
>> - I think I understand this, but it was not an easy read.
>> ---
>> End of section 3, first bullet, para 2:
>> /dangerously discarding/
>> - I agree that this results in discard of end-to-end congestion signals,
>> and that from a network stability perspective this is undesirable, but I
>> think the word "dangerous" needs more explanation.
>> ---
>> After Fig 3:
>> /as the ingress end of RFC4301 IPsec/
>> - Could this be written better, as something like:
>> /as the tunnel ingress processing of IPsec in RFC4301/
>> ---
>> /  Not only will this streamline PCN
>>       standardisation, but it could also streamline other future uses of
>>       these codepoints./
>> - What exactly does this mean?
>> ---
>> End of 4.2:
>> /not so often that the illegal combinations would amplify into a flood
>> of alarm messages/
>> - I agree that these need to be controlled. Can you say "rate-limited" -
>> is there some more helpful bound we can cite here for guidance?
>> ---
>> Section 4.3:
>> /S.5/
>>  ^ - Section?
>> -- 
>> Section 4.3, bullet 1
>> /should not set the ECN field/
>>             ^^^
>> - Do you mean set this to CE? or do you mean re-mark? Please be more
>> specific.
>> ---
>> Section 5:
>> /Note: in/
>>  ^^^^^^^
>> - Would it be OK to say "In"?
>> ---
>> Section 5, para 1:
>> /doesn't need/
>>     ^^^ - could we say /does not need/
>> ---
>> Section 5, para 1:
>> /to know what mode the ingress is in./
>>     ^^^^^                      ^^^^^
>> - Could this be /need to determine the mode of the tunnel ingress/.
>> ---
>> Section 5.1, para 2:
>> /whichever mode the tunnel was in/
>>                            ^^^^^^
>> - could you improve the English of this line.
>> ---
>> Section 5.1, para 3:
>> /as its behaviour/
>>  ^^^^^^
>> /since the behaviour/
>> ---
>> Section 5.1, final para:
>> /doesn't/ and /don't/
>> - could you say /does not/ and /do not/
>> ---
>> Section 4.3, title:
>> /Upgrading/
>>  ^^^^^^^^
>> - Is this upgrading or updating, I'd prefer the latter if you think this
>> is applicable.
>> ---
>> Section 5.3, para 1:
>> /dangerously discarding/
>> - I agree that this results in discard of end-to-end congestion signals,
>> and that from a network stability perspective this is undesirable, but I
>> think the word "dangerous" needs more explanation.
>> ---
>> Section 5.3, para 1:
>> /ratchet up/ and /ratchet down/
>> - I agree, but this could be written more clearly.
>> ---
>> Section 5.3, para 4:
>> /if the egress says/
>>                ^^^
>> - advertises? - can you select a better word?
>> /doesn't understand the question/
>> - could you rephrase this please in terms of the protocol or set of
>> actions that need to take place?
>> ---
>> Section 6, para 1:
>> /is a change to the ingress/
>>     ^^^^^^^^
>> - is it better to say /updates the ingress/?
>> ---
>> References:
>> draft-ietf-pcn-architecture has now been published as RFC 5559.
>> ---
>>
> 
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research
> 



From rbriscoe@jungle.bt.co.uk  Wed Jul  8 05:46:38 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E94173A6F5C for <tsvwg@core3.amsl.com>; Wed,  8 Jul 2009 05:46:38 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.943
X-Spam-Level: 
X-Spam-Status: No, score=-1.943 tagged_above=-999 required=5 tests=[AWL=0.174,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id TELrtcaQFUKu for <tsvwg@core3.amsl.com>; Wed,  8 Jul 2009 05:46:32 -0700 (PDT)
Received: from smtp3.smtp.bt.com (smtp3.smtp.bt.com [217.32.164.138]) by core3.amsl.com (Postfix) with ESMTP id AC50A3A6F63 for <tsvwg@ietf.org>; Wed,  8 Jul 2009 05:46:31 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp3.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Wed, 8 Jul 2009 13:43:23 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Wed, 8 Jul 2009 13:43:23 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247057001449; Wed, 8 Jul 2009 13:43:21 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n68ChEUm026763; Wed, 8 Jul 2009 13:43:14 +0100
Message-Id: <200907081243.n68ChEUm026763@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Wed, 08 Jul 2009 13:43:18 +0100
To: "Ingemar Johansson S" <ingemar.s.johansson@ericsson.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <130EBB38279E9847BAAAE0B8F9905F8C0174F1E7@esealmw109.eemea. ericsson.se>
References: <mailman.11454.1246022939.4936.tsvwg@ietf.org> <130EBB38279E9847BAAAE0B8F9905F8C0174F1E7@esealmw109.eemea.ericsson.se>
Mime-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"; format=flowed
Content-Transfer-Encoding: quoted-printable
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 08 Jul 2009 12:43:23.0055 (UTC) FILETIME=[AE0FCBF0:01C9FFC9]
Cc: Ingemar Johansson S <ingemar.s.johansson@ericsson.com>, tsvwg@ietf.org
Subject: Re: [tsvwg] Comments on draft-ietf-tsvwg-ecn-tunnel-02
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 08 Jul 2009 12:46:39 -0000

Ingemar,

At 20:40 29/06/2009, Ingemar Johansson S wrote:
>Hi
>
>I have read draft-ietf-tsvwg-ecn-tunnel-02 and below some comments.
>
>Figure 4:
>  I wonder what will happen if we have a TCP=20
> stack that implements ECN-nonce. Wouldn't the=20
> rules in figure 4 lead to a faulty nonce-sum=20
> with negative consequences for the TCP throughput. Or is this out of=
 scope?.

Your point is certainly in scope, but I don't=20
believe the criticism is correct. The transitions=20
were designed to allow the nonce to work (in fact better than it did=
 before).

If routers leave arriving ECT(0) or ECT(1)=20
unchanged (as they should) the tunnel egress=20
behaviour defined in Table 4 will also preserve=20
ECT(0) & ECT(1). There will be no faulty nonce sum, and all will be fine.

The nonce is designed to detect a change of=20
ECT(0) to ECT(1) or vice versa. The old (RFC3168)=20
egress behaviour undid any change within a tunnel=20
between either of these codepoints, so it=20
couldn't be detected outside the tunnel. We've=20
changed the egress behaviour so at least a change=20
in one direction can be detected outside the=20
tunnel. The security considerations section=20
explains why we decided it was best to leave a=20
change in the other direction undetectable (as it had been in RFC3168).


>Page 19: "sfaeguards" -> "safeguards"
>
>Page 22 bullet 4. (ECN nonce) "..no evidence=20
>that anyone has implemented it beyond the=20
>author's prototype". This one made me curios...=20
>In general one get the feeling that it is not=20
>good or recommended to use ECT(1) e.g as a light=20
>congetsion indication (I believe it is written=20
>in similar words in RFC3168), the main reason is=20
>guess the ECN-nonce, this bullet says "don't=20
>care about the ECN-nonce". Is there any=20
>reference to a deeper discussion on around this?

RFC4774 gives guidelines for the possibility of=20
multi-level ECN semantics. The behaviour in Table=20
4 is designed to allow both Nonce-like behaviour=20
and multi-level ECN behaviour to work, whereas in RFC3168 & RFC4301 broke=
 both.

This bullet doesn't say "Don't use the ECN nonce"=20
at all. RFC3168 and RFC4301 were both worse for=20
the ECN Nonce. We've made them better in one=20
case, but we couldn't in the other.

Specifically, RFC3168 broke the function of the=20
ECN nonce for detecting changes between ECT(0)=20
and ECT(1), but only changes within a tunnel. The=20
Nonce worked and Table 4 allows it to still work=20
for detecting changes by the receiver - it always=20
did and still does. We've managed to at least=20
make the Nonce work for a change in one=20
direction, but fixing it for the other direction=20
prevented use of multi-level ECN. This simply=20
said, "We decided the corner-case of detecting a=20
network operator suppressing congestion=20
notification within a tunnel isn't as important=20
as detecting cheating receivers or multi-level ECN."

The case left broken is such a corner case, I can=20
see I'm going to have to make it clear that the=20
Nonce still works in the normal cases, and in one more case than it did=
 before.


>Page 29: "DIffserv" -> "Diffserv"

Tx for the comments (and the nits). I'll issue a=20
new draft this week (also taking account of=20
Gorry's longer review, which I will respond to shortly).

Cheers


Bob


>Regards
>Ingemar
>*******************************************
>Ingemar Johansson
>Senior Research Engineer, IETF "nethead"
>EAB/TVK - Multimedia Technologies
>Ericsson Research Ericsson AB
>Box 920 S-971 28 Lule=E5, Sweden
>Tel: +46 (0)10 7143042
>ECN: 852-43042
>ECC: 852-19042
>Mobile: +46 (0)730 783289
>Visit http://labs.ericsson.com !
>*******************************************

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research=20


From root@core3.amsl.com  Wed Jul  8 09:45:02 2009
Return-Path: <root@core3.amsl.com>
X-Original-To: tsvwg@ietf.org
Delivered-To: tsvwg@core3.amsl.com
Received: by core3.amsl.com (Postfix, from userid 0) id 06F763A69EB; Wed,  8 Jul 2009 09:45:01 -0700 (PDT)
From: Internet-Drafts@ietf.org
To: i-d-announce@ietf.org
Content-Type: Multipart/Mixed; Boundary="NextPart"
Mime-Version: 1.0
Message-Id: <20090708164502.06F763A69EB@core3.amsl.com>
Date: Wed,  8 Jul 2009 09:45:02 -0700 (PDT)
Cc: tsvwg@ietf.org
Subject: [tsvwg] I-D Action:draft-ietf-tsvwg-dtls-for-sctp-01.txt
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 08 Jul 2009 16:45:02 -0000

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Transport Area Working Group Working Group of the IETF.


	Title           : Datagram Transport Layer Security for Stream Control Transmission Protocol
	Author(s)       : M. Tuexen, et al.
	Filename        : draft-ietf-tsvwg-dtls-for-sctp-01.txt
	Pages           : 9
	Date            : 2009-07-08

This document describes the usage of the Datagram Transport Layer
Security (DTLS) protocol over the Stream Control Transmission
Protocol (SCTP).

The user of DTLS over SCTP can take advantage of all features
provided by SCTP and its extensions, especially support of

o  multi-homing to provide network level fault tolerance.

o  multiple streams to avoid head of line blocking.

o  unordered delivery.

o  dynamic reconfiguration of streams.

o  partially reliable data transfer.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-tsvwg-dtls-for-sctp-01.txt

Internet-Drafts are also available by anonymous FTP at:
ftp://ftp.ietf.org/internet-drafts/

Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Message/External-body;
	name="draft-ietf-tsvwg-dtls-for-sctp-01.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID: <2009-07-08093157.I-D@ietf.org>


--NextPart--

From rbriscoe@jungle.bt.co.uk  Thu Jul  9 02:53:45 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 319B03A6CC8 for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 02:53:45 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.993
X-Spam-Level: 
X-Spam-Status: No, score=-1.993 tagged_above=-999 required=5 tests=[AWL=0.124,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id HNNdeaBmuuyb for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 02:53:39 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id 364FC3A6A93 for <tsvwg@ietf.org>; Thu,  9 Jul 2009 02:53:38 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 9 Jul 2009 10:54:05 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Thu, 9 Jul 2009 10:54:05 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247133244984; Thu, 9 Jul 2009 10:54:04 +0100
Received: from mut.jungle.bt.co.uk ([10.73.213.20]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n699rvdt027065; Thu, 9 Jul 2009 10:53:57 +0100
Message-Id: <200907090953.n699rvdt027065@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Thu, 09 Jul 2009 10:54:07 +0100
To: gorry@erg.abdn.ac.uk, Michael Menth <menth@informatik.uni-wuerzburg.de>, "MONCASTER, Toby" <toby.moncaster@bt.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <4A52F472.3040500@erg.abdn.ac.uk>
References: <4A43D532.6040906@erg.abdn.ac.uk> <200906261328.n5QDSuiM007908@bagheera.jungle.bt.co.uk> <4A52F472.3040500@erg.abdn.ac.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 09 Jul 2009 09:54:05.0543 (UTC) FILETIME=[32203770:01CA007B]
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] Some comments on draft-ietf-tsvwg-ecn-tunnel-02]]
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 09 Jul 2009 09:53:45 -0000

Gorry, Michael, Toby,

I'm dealing with all your detailed comments by a rev of the draft.

I'll respond to each comment on the list once I've worked out better 
text to address all your comments in one go.

In general, I think the main problem with my text is that I've not 
stated enough background on how things work (or at least how I assume 
things work), as tunnelling is a subject you really have to think 
hard about. In Lars's words "This tunnelling stuff makes my head hurt."

Just one response to Gorry's top level comments below...

At 08:08 07/07/2009, Gorry Fairhurst wrote:
>Date: Tue, 07 Jul 2009 08:08:34 +0100
>From: Gorry Fairhurst <gorry@erg.abdn.ac.uk>
>To: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
>CC: tsvwg <tsvwg@ietf.org>
>Subject: Re: [tsvwg] Some comments on draft-ietf-tsvwg-ecn-tunnel-02]]
>
>See a few responses to your comments in-line,
>
>Gorry
>
>Bob Briscoe wrote:
>>Gorry,
>>Tx for such a thoro review. It seems nearly all the comments are 
>>about poor description, rather than any protocol disagreements. Is this fair?
>Yes.
>
>>The only substantive disagreement seems to be over whether ECT(1) 
>>encapsulating  CE should raise an alarm at the egress.
>I am assuming that you think the role/use of tunnel negotiation is 
>also a document organisation issue?

Yes, I think your concerns over tunnel negotation are to do with my 
bad explanation, rather than protocol disagreement.


Bob


________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From toby.moncaster@bt.com  Thu Jul  9 06:42:53 2009
Return-Path: <toby.moncaster@bt.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 2B7243A6A32 for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 06:42:53 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.68
X-Spam-Level: 
X-Spam-Status: No, score=-2.68 tagged_above=-999 required=5 tests=[AWL=-0.281,  BAYES_00=-2.599, J_CHICKENPOX_72=0.6, J_CHICKENPOX_91=0.6, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id pbTm8nEU+vnB for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 06:42:52 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id EA8813A6920 for <tsvwg@ietf.org>; Thu,  9 Jul 2009 06:42:50 -0700 (PDT)
Received: from E03MVZ1-UKDY.domain1.systemhost.net ([193.113.30.63]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 9 Jul 2009 14:43:17 +0100
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Thu, 9 Jul 2009 14:42:59 +0100
Message-ID: <AEDCAF87EEC94F49BA92EBDD49854CC70C239DDF@E03MVZ1-UKDY.domain1.systemhost.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: comments on draft-ietf-tsvwg-ecn-tunnel
Thread-Index: Acn18Z4/bFVGTrNAQ9C4cVWL3QGdAQCC89WAAidiYbA=
From: <toby.moncaster@bt.com>
To: <tsvwg@ietf.org>
X-OriginalArrivalTime: 09 Jul 2009 13:43:17.0662 (UTC) FILETIME=[37070BE0:01CA009B]
Subject: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 09 Jul 2009 13:42:53 -0000

Better copy these comments to the working group!

Toby

-----Original Message-----
From: Moncaster,T,Toby,CXR9 R=20
Sent: 01 July 2009 17:43
To: Briscoe,RJ,Bob,XVR9 BRISCORJ R
Subject: RE: Calling in your offer to review draft-ietf-tsvwg-ecn-tunnel

Hi Bob,

I probably was silly enough to volunteer! I also remember volunteering
to review a couple of other IETF things back in March that are no doubt
going to catch up with me soon! Will do my best to make this a useful
(e.g. constructive) review...

Abstract:

Seems to lack detail of what you are expecting people to do having read
this. Need to clarify that whilst this updates RFC4301 & RFC3168 it only
ADDS functionality to 4301 (in other words 4301 compliant endpoints will
always be a subset of those compliant with this standard).

Introduction:

Overall - seems a bit over-long and gives too much detail both about
WHAT you are intending to do and WHY. I would suggest summarise the what
and concentrate on the why (as that is going to get less explanation in
the rest of the document).

Make it clear from the start that you are trying to impose a common set
of rules on this. Something like "currently the behaviour of IP-in-IP
tunnel endpoints is not consistent, with consequent risk of
misconfigurations etc etc..."

Make it more obvious that the changes on entry are to bring all current
behaviours into line. And thus that the major changes are at the
egress...

Re-cap of ECN is irrelevant for this document - you can sum this up in 1
sentence here and move the table into a later section or an appendix...

You still seem to have too much justification text explaining why this
situation occurred - this could be fine later on but is too detailed for
the intro.

" As a result, an unfortunate sequence of standards actions has left us
with nearly the worst of all possible combinations of outcomes, despite
the best endeavours of everyone concerned."
Sounds political to me (your telling them they screwed up but you try to
sugar the pill). Only really need to say that we have ended up with
contradictions (and I believe it is more that the different groups were
trying to second-guess each other rather than because they were
expending their best endeavours).

You mention PCN twice - first as the reason for needing to do something
now then as the main motivator for doing anything at all. I would
suggest stick to the first take - PCN highlights the absurdity of the
current situation so now is the time to do something. However longer
term this is going to sound strange once PCN is standardised and this is
accepted so it might be worth pointing out this text is not relevant to
the standard itself?

1.1 Scope - this seems fine.=20

1.2 Document Roadmap seems fine (but hides the presence of many more
appendixes)

Section 3.

Could this become mainly Appendix material? Keep a summary, but cut the
length? Are you effectively defining the term "resetting" an ECN
codepoint? If so have a definitions section...

"But if non-IPsec tunnels are already present within a network to which
PCN is being added, that is not particularly helpful advice." - indeed,
but it is exactly as unhelpful to define a new set of standards
behaviours!!!

Section 3.2 can certainly be shorter...

Section 4 is the main point of the document... Make it clear this is the
case

In figure 4 include a footer describing the meaning of (!!!) as some
people will fial to read through the actual text...

4.1

"Compatibility mode is the same per packet behaviour as the ingress end
of RFC3168's limited functionality mode." --> Compatibility mode behaves
the same as the limited functionality mode in RFC3168 but only applies
at the tunnel ingress.

4.2
First bullet below figure 4 "If the inner ECN field is Not-ECT the
decapsulator MUST NOT propagate any other ECN codepoint in the outer
header onwards." --> If the inner ECN field is Not-ECT the decapsulator
MUST NOT propagate any other ECN codepoint forwards.

First sub-bullet below figure 4 - I believe this may be actually defined
in RFC4774 as the only safe behaviour. Probably worth spelling out that
the issue is that the congestion mark may not get reported back to the
sender.

3rd bullet - would be nice to explain the SHOULD (do you log every time,
or just log that it has happened; what circs is it ok not to log these)

As section 4.3 is informative it would be clearer to move it into its
own section (so all the normative text is in a single place).

End of 1st proper para: "Therefore here we give guidance for designing
possibly different marking schemes." Is very clunky... do you mean
guidance for designing different marking schemes, possible different
marking schemes, marking schemes that possibly differ,...

First bullet: xref to RFC4774 which has strong parallels with this
advice. I know 4774 is badly written but people are going to notice that
there is overlap between these and so it would be good to reference it
(e.g. "RFC4774 provides guidance on overloading the ECN field...")

3rd bullet - the only instance this bullet actually covers is that of
not ECT on the inner header so why not just say this?

5. End of 1st para. Highlight that this is a big advantage as it is
likely to reduce possible misconfigurations

Near the end you have some normative text - this is adrift of the main
block of such text so should actually be in section 4 (although the
explanations for it can be here).


Section 6: can this be presented in concise bullets at the start of the
section? E.g. this updates 3168 in following ways... etc

Section 8 - I suggest having the details of the Nonce hole put into an
Appendix with an xref from here...

Update references especially RFC5559 PCN!

Appendixes - I take it eventually these will get significantly reduced
and are really there for the purposes of explanation during
standardisation? The following comments are if you are going to keep the
appendixes...

App A.1 This seems to be a mixture of security 101 (as the Americans
would put it) and specific stuff for this actual work. Do you need all
the tutorial stuff?

A.2 Sure this could be a bit more concise!

A.3 This seems the right sort of level of detail

B.1 can any of this be explained by a suitable reference?

B.2 & B.3 Seem excessively detailed given you now have more or less
convinced people this is worthwhile work.

Appendix D title is slightly ambiguous (you have to work out where the
emphasis goes...).


HTH

Toby


From rbriscoe@jungle.bt.co.uk  Thu Jul  9 12:07:50 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E05F93A67DF for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 12:07:50 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.008
X-Spam-Level: 
X-Spam-Status: No, score=-3.008 tagged_above=-999 required=5 tests=[AWL=1.109,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 2q1s6Oh3IV6Y for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 12:07:44 -0700 (PDT)
Received: from smtp4.smtp.bt.com (smtp4.smtp.bt.com [217.32.164.151]) by core3.amsl.com (Postfix) with ESMTP id 4E09A28C15D for <tsvwg@ietf.org>; Thu,  9 Jul 2009 12:07:44 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp4.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 9 Jul 2009 20:08:11 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Thu, 9 Jul 2009 20:08:11 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247166490419; Thu, 9 Jul 2009 20:08:10 +0100
Received: from mut.jungle.bt.co.uk ([10.73.213.20]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n69J838L011064; Thu, 9 Jul 2009 20:08:03 +0100
Message-Id: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Thu, 09 Jul 2009 20:08:07 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 09 Jul 2009 19:08:11.0260 (UTC) FILETIME=[9A1E0BC0:01CA00C8]
Cc: gorry@erg.abdn.ac.uk, tsvwg IETF list <tsvwg@ietf.org>
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 09 Jul 2009 19:07:51 -0000

David,

Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...

Your comments get to the nub of the proposed egress protocol changes. 
Great... inline...

At 00:43 09/07/2009, Black_David@emc.com wrote:
>Bob,
>
>Sorry for the delay on this: things got entirely too busy around the
>end of the quarter here.  Having re-read the draft from scratch, I
>have a number of things to comment on ... I will be in Stockholm,
>and some of these items are going to need further discussion
>(private and/or public) in Stockholm.
>
>Starting with this table in Figure 4:
>
>+------------------+----------------------------------------------+
>|  Incoming Inner  |             Incoming Outer Header            |
>|      Header      +---------+------------+------------+----------+
>|                  | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
>+------------------+---------+------------+------------+----------+
>|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)| drop(!!!)|
>|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |   CE     |
>|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |   CE     |
>|        CE        |      CE |     CE     |     CE(!!!)|   CE     |
>+------------------+---------+------------+------------+----------+
>                    |                Outgoing Header               |
>                    +----------------------------------------------+
>
>My initial focus is on what needs to change from RFC 3168/RFC 4301
>behavior and why.  Given that these two RFCs line up, I think changes
>need strong motivation.  So, here are the four changes, and my thoughts:
>
>a) Inner: Not-ECT, Outer: ECT(0)
>         3168/4301: drop (!!!)
>         Change: Not-ECT (!!!)
>         Rationale: "allow future standards actions to use this
>                 combination"
>         My view: This ain't broke, why are you fixing it?

It is broke in the sense tunnel egresses unnecessarily drop harmless 
packets merely because they carry a "currently unused" (CU) 
combination of codepoints. This is similar to the problem ECN faced 
with firewalls dropping packets unless they were Not-ECT.

(a) wasn't originally part of the changes I was proposing. But Anil 
Agawaal convinced me it is worth doing. Reasoning: If we are making a 
change to decap anyway, let's get it as right as possible. If we 
decide not to make any other changes to decap, I wouldn't fight for 
this one alone.


>b) Inner: ECT(0), Outer: ECT(1)
>         3168/4301: ECT(0)
>         Change: ECT(1)
>         Rationale: Treat ECT(1) as more severe congestion than ECT(0).
>         My view: Longer discussion below.
>
>c) Inner: ECT(1), Outer: ECT(0)
>         3168/4301: ECT(1)
>         Change: ECT(1) (!!!)
>         Rationale: "no currently legal transition" can produce this
>combo.
>         My view: Not a good idea, among other reasons because it's at
>                 odds with your "future standards action" explanation for
>                 the first change.  I'd remove this change and instead
>deal
>                 with the possibility of raising an alarm in a
>PCN-specific
>                 draft rather than declaring this to be an alarm-able
>combo
>                 in general.
>
>d) Inner: CE, Outer: ECT(1)
>         3168/4301: CE
>         Change: CE (!!!)
>         Rationale & My view: <same as previous change.

c) & d) aren't PCN specific.

Reasoning for adding alarms (also said this eariler on list in 
conversation with Gorry):
If security folks have concerns over possible attacks from "currently 
unused" combinations, I want to give them a way to manage the 
perceived threat without just dropping packets. RFC3168 recommended 
alarms for CU combinations. In these cases, I believe RFC3168 simply 
overlooked that these are also CU combinations.


>So, I would remove a) as not defensible (no good reason for change),
>and I would push the discussion of adding alarms for c) and d) into
>a PCN-specific document.

c) & d) are not to do with PCN.

My motivation is: While making change (b) to enable PCN, lets do 
(a,c,d) to make the IP layer as clean as possible, to ensure easier 
future evolvability.

I don't see the issue. It says SHOULD log and MAY alarm. So the 
operator doesn't have to do anything.

>The c) alarm in particular is only
>meaningful if PCN or something like it is in use.

No. The combination in c) always was CU, and it's still CU with PCN. 
None of the proposed changes are PCN-specific.

Even (b) is intended to also be useful for any 3-level ECN while not 
precluding 2-level ECN and the nonce.

NB:  3 level = 2 'congested' levels + 1 level not congested, e.g. VCP 
or PCN which were both mentioned in RFC4774. This doesn't count the 
not-ECN-capable state, which makes 4 states.

>I don't view d)
>as deserving of an alarm in general, as even though something has
>happened, no damage has been done - this is in line with the
>"be liberal in what you accept" principle of protocol design.

That's why I'm proposing an optional alarm. Optional alarms *are* 
liberal, but they allow operators a way to address security concerns 
about 'known unknowns' [Rumsfelt], while still allowing new 
behaviours to be introduced.


>That leaves b), which I think is the key desired change, because
>it makes a third ECN code point usable through tunnels.

Yes - exactly.

>Much as
>this might be useful for PCN, it may be too late to do this.  The
>problem starts from the need to say that this draft updates RFC 4301.
>The result will be a mix of 4301 and updated-4301 implementations
>that render the third state difficult to use in practice.  The
>problem's even worse for non-IPsec tunnels.

The idea is that an operator wanting to implement PCN will have a 
spec they can point to in their invitation to tender. At the same 
time this spec both UPDATEs and is backward compatible with RFC3168 & 
RFC4301 because it only changes CU stuff at egress and the change at 
the ingress leads to the same output at the egress. Therefore vendors 
can update 3168 & 4301 tunnels lazily - only when their customers ask 
for the new features (whether for PCN or for something else like VCP).

>I also read Appendix D
>as saying that PCN is going the two-state route, not three.

That's not how it should be read - I'll clarify it. The PCN w-g wants 
3 states, but can't have them while ecn-tunnel is not proposed 
standard. Therefore, it has decided to define a 2-state baseline and 
a number of experimental schemes to work round this tunnel issue 
(plus a comparison doc about them all). It has agreed that eventually 
just one experimental scheme will be chosen for standards track:
- 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
- 3-state (which uses an extra DSCP)
- psdm (which requires out of band state)
- sm (single marking gets 3 functions from 2 encoding states but 
admits admission control & termination aren't very precise in all cases)

___
See part 2/2 for responses to your comments on specific parts of the test.

Thanks again for the review


Bob



________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research  


From rbriscoe@jungle.bt.co.uk  Thu Jul  9 12:14:50 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 125D73A6D2E for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 12:14:50 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.131
X-Spam-Level: 
X-Spam-Status: No, score=-2.131 tagged_above=-999 required=5 tests=[AWL=-0.014, BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id ziG+dSTUGRSa for <tsvwg@core3.amsl.com>; Thu,  9 Jul 2009 12:14:43 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id 9C99B3A6808 for <tsvwg@ietf.org>; Thu,  9 Jul 2009 12:14:02 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 9 Jul 2009 20:14:29 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Thu, 9 Jul 2009 20:14:29 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247166868662; Thu, 9 Jul 2009 20:14:28 +0100
Received: from mut.jungle.bt.co.uk ([10.73.213.20]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n69JEM4Y011182; Thu, 9 Jul 2009 20:14:23 +0100
Message-Id: <200907091914.n69JEM4Y011182@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Thu, 09 Jul 2009 20:14:28 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 09 Jul 2009 19:14:29.0160 (UTC) FILETIME=[7B5D0280:01CA00C9]
Cc: gorry@erg.abdn.ac.uk, tsvwg IETF list <tsvwg@ietf.org>
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 2/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 09 Jul 2009 19:14:50 -0000

David,

Part 2/2...

Thanks for the further comments below. I agree with everything and 
will make appropriate changes, except where I've pushed back below.

At 00:43 09/07/2009, Black_David@emc.com wrote:

[snip egress changes stuff - see previous mail]

>To counter this, the argument in Appendix D) needs to be strengthened;
>I think there's an appeal to the controlled domain/closed system
>nature of PCN - i.e., PCN is only used by an operator who is in
>complete control of everything in the PCN domain, hence can assure
>that all tunnels are 3-ECN-value-compatible.  You haven't said this,
>you need to, and probably in the main body of the draft, not hiding
>in Appendix D).

Yup.


>Section 4.3 should be an Appendix, with the possible exception of
>the paragraph containing the "Don't" guidance.

I don't think moving this to an appendix is the appropriate way to 
deal with future design guidelines. My intent here was to encapsulate 
a tiny informational RFC within a standards track RFC. Gorry went the 
other way to you and asked whether some of the shoulds and should 
nots would be better capitalised.

I need guidance here, as I don't know what is normal procedure when 
including guidelines for future protocol design in a stds track doc. 
I noticed Fred's admitted-realtime-dscp contains a section with 
uncapitalised recommendations, and I figured this is more useful than 
a tiny separate I-D, which would have to "re-load all the state" 
built up within ecn-tunnel.

Possibilities:
1. Drop the whole section
2. Keep it as it is
3. Write a short informational RFC just for this section.
4. Move the bullet points to an appendix, but keep a short guidance 
para in the body.

My position on these:
1. No: ecn-tunnel has achieved an ECN tunnelling behaviour that works 
across a wide range of requirements. This section gives sound advice 
on trying to keep it that way. There is no other better place for 
this advice than in ecn-tunnel, because RFC4774 (alt ecn sematics) 
doesn't cover tunnels.

2. Yes. I think it's fine as it is. I'll sign-post it so implementers 
can skip over it. I'll also cite RFC4474.

3. No. A new I-D would need loads more contextual text copied from 
this one than the actual short piece of guidelines text.

4. May-be. But that just makes the whole doc longer with redundant 
headings and cross-refs and it reduces the force of the guidelines.


>Section 5.2 fails to discuss the 3rd-value incompatibility problems
>caused by a non-upgraded (i.e., 4301-only) egress.  It's correct
>that the 4301-only and new egress behaviors are both compatible
>with a 4301 ingress.  In essence, 5.2 has failed to discuss egress
>behavior changes.

5.2 is about ingress back-compat. 5.1 is about egress, and I believe 
covers what you are asking for.


>The last few paragraphs of Section 5.3 that say it's not enough to
>just implement compatibility mode are going to cause problems

Not at all. This says what an implementor has to do to comply with 
this spec. It says an ingress that always turns off ECN processing 
across a tunnel can't be considered compliant with the ECN-tunnelling spec.

IOW, if a vendor doesn't want to support ECN tunnelling at all, they 
just won't comply with this spec. It doesn't mean they have to 
support ECN tunnelling, it just means they can't claim they comply 
with the ECN-tunnelling spec if they don't.

Surely this is all just stating the obvious.

>(in
>essence, a lot of existing tunnels immediately become non-compliant).

Surely legacy stuff is always non-compliant with new RFCs that update 
the RFC they were coded against. E.g. an RFC2003 tunnel is 
non-compliant with RFC3168. The ingress ECN processing of an 
RFC3168/2401 IPsec tunnel is non-compliant with RFC4301.

>I got a private email from Gorry on basically this point, in which
>Gorry essentially asks (I've added emphasis): How on earth is an
>existing tunnel supposed to negotiate this?

I missed out saying (and I'll add) that I'm not recommending 
negotiation, but I have to recognise that it may exist (because 3168 
introduced it). There are two potential types of tunnel:
* Static: those coded on the assumption that the pair of ends are 
config'd together by the same admin
* Adaptive: those where each end is coded to be able to work with 
whatever other end it encounters as it sets up tunnels driven by the 
demands of arriving packets. E.g. Dynamic TED, which isn't IETF 
standardised AFAIK, but it exists: 
<http://www.cisco.com/en/US/docs/ios/12_0t/12_0t5/feature/guide/ted.html>

Existing static tunnels won't have any problem with a new spec; if 
the admin upgrades one, it's up to the admin to upgrade the other at 
the same time.

If an existing adaptive tunnel already negotiates between RFC3168 
limited & full functionality, this spec is designed so a new ingress 
or a new egress "just works" with such a negotiating legacy tunnel endpoint.

>Another way to realize
>that you have a problem here is that you also need to say that
>this draft updates RFC 2003, and moreover does so in a non-backwards-
>compatible fashion.  I'm not going to say it has to come out, but
>attention clearly needs to be called to this in Stockholm.

I say "This specification updates RFC3168 and RFC4301."
And I say "In the RFC index, RFC3168 should be identified as an 
update to RFC2003. RFC4301 should be identified as an update to RFC3168."

Therefore, as this spec updates 3168, which updates 2003, this spec 
updates 2003.

But updating 2003 is not a problem. This new spec is designed to 
ensure a compliant tunnel endpoint will interwork safely with a 2003 
tunnel endpoint at the other end. Whichever end is which.

For the full scary table of all the possible permutations of 
compatibility, see spare slide #11 in the pack I presented at the last IETF:
<http://www.bobbriscoe.net/present.html#0903ecn-tunnel>
As you can see, the new spec interworks with all the possible other 
tunnel endpoints.


>Section 6 needs to deal with RFC 2003.

OK, that's easy to do (I was merely trying to avoid extra text about 
double backwards compatibility, but it's been considered in the 
wording of the rules so it's no problem to explain this).


>Somewhere in Section 5.3 and 6, you ought to discuss RFC 3168's
>update to RFC 2401.  To put it bluntly, this was a failed attempt
>to add ECN egress behavior negotiation to an existing tunnel
>setup protocol (IKEv1 for a tunnel-mode SA).  The result was
>unworkable in practice for overhead and negotiation structure
>reasons, and I suspect it was never implemented.

I'd like to know more about the history here.

By my analysis, an RFC3168 _egress_ never needed to know the mode the 
tunnel was in, because a full mode egress produces the correct output 
in either case:
- limited mode ingress:   outer    = Not-ECT
- limited mode egress:    outgoing = inner
- full mode egress:       outgoing = inner (if outer = Not-ECT)

Before entering full mode, an RFC3168 (or later) _ingress_ does need 
to know it's not talking to a pre-3168 egress which would discard CE 
introduced within the tunnel.

If this implies I need to define how an ingress checks the egress is 
ECN-capable, please tell all!


>Section 8 spends too much text on the ECN nonce.

I agree. The nonce issues within tunnels are a really unimportant 
aspect of the nonce.

But if I don't talk about corner-cases, I'll only get someone on the 
IESG showing off that they've noticed I haven't taken account of a 
corner-case ;)

It's the same with the IKEv2 vs manual keying corner case.

I'll try to say less, but it's hard, because the corner-cases require 
convoluted description of what the case is.

As this is a change to the IP layer, I do feel I have to say at least 
something about anything that it might impact.

>As part of
>discussion in Stockholm, I'd ask about whether this draft should
>be used as an opportunity to declare the ECN nonce (RFC 3540)
>to be Historic.

I don't believe the nonce should be historic until something else 
(e.g. re-ECN) looks like it will replace it. I think it would be 
previous to make it historic otherwise - it's not wrong at all, it's 
just the best that could be thought of at the time.

The nonce is a fine idea, it's just too limited in applicability IMHO 
to be in the IP layer. The nonce only works for e2e transports with 
in-order reliable delivery, and it is only applicable if the sender 
is expected to act in the interests of the common good.

My only concern is if anyone actually implemented the nonce it would 
make re-ECN hard/impossible to deploy. Because with re-ECN a network 
operator is intended to count ECT(1) packets without looking at flows 
(and in re-ECN, ECT(1) has a different meaning to the nonce).


>On rereading this draft from scratch, I'm not clear on why
>Appendix B is included.  I think Section A.2 of Appendix A
>is sufficient (or if not, can be expanded slightly).  Appendix
>B is very abstract, and needs better grounding in actual examples
>(e.g., not saving them until the end).  I'd remove Appendix B.

I've just re-read it and you're right. It can be removed. Others 
wanted shoter appendices, so this solves that problem in a stroke!


>The paragraph in Section A.2 on egress congestion differentiation
>between congestion across a tunnel vs. upstream.  In contrast
>to Appendix B, I think Appendix C is ok in its current form, I
>just don't see the need for it and would remove it.

I'm confused. Do you mean remove appendix B (which I agree with)? Or 
remove appendix C (which I'd rather not)?

Appx C addresses the only concern anyone could have about losing the 
reset behaviour of a 3168 ingress.


>There's some Appendix D discussion above.  In addition to that,
>D.1 needs to be reworked to better acknowledge the current
>2-state PCN direction.

Well, I agree it needs to be updated to acknowledge the current state 
of discussions in PCN (but this isn't just a 2-state direction as I 
explained above).


>Last paragraph in Appendix E needs a little more text to explain
>that what goes wrong is (wrt Fig 6) that "R" marks some traffic
>for removal by setting CE, and then "M" independently marks the
>same fraction of traffic for removal - if there's no overlap in
>the markings, twice as much traffic gets removed as should have
>been.

It's not like this, but this highlights to me that I need to re-write 
that para to explain it better to those that don't know PCN intimately.


>In addition, discussion of this scenario (too much traffic removed)
>probably needs to say something about how/why this can still happen
>despite the controlled domain/closed system nature of PCN (which I
>think needs to be added - see above).

OK.

Thanks v much for this excellent & thoro review. I'm revving -02 to -03 now.


Bob


>Enjoy ;-),
>--David
>
> > -----Original Message-----
> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > Sent: Thursday, June 25, 2009 8:07 PM
> > To: Black, David
> > Subject: Calling in your offer to review draft-ietf-tsvwg-ecn-tunnel
> >
> > David,
> >
> > I noted down that you volunteered to review this one, in a weak
> > moment at an earlier IETF...
> > <http://tools.ietf.org/wg/tsvwg/draft-ietf-tsvwg-ecn-tunnel/>
> >
> > I'd appreciate comments by Mon 6 Jul '09 if poss - when I return from
> > a week off so I can make revs for the IETF deadline the
> > following week.
> >
> > I've substantially re-written it - simplified and cut-down. As with
> > many drafts after a few revs, it needed a complete re-write - all the
> > modifications had got unwieldy and in hindsight most of the
> > justificatory text could be removed to appendices, leaving just the
> > standards action stuff.
> >
> > Cheers
> >
> >
> > Bob
> >
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research  


From rbriscoe@jungle.bt.co.uk  Fri Jul 10 06:11:49 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 189F528C2EE for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 06:11:49 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.634
X-Spam-Level: 
X-Spam-Status: No, score=-1.634 tagged_above=-999 required=5 tests=[AWL=-0.509, BAYES_00=-2.599, DATE_IN_PAST_12_24=0.992, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id dFeb7YFQQso4 for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 06:11:43 -0700 (PDT)
Received: from smtp3.smtp.bt.com (smtp3.smtp.bt.com [217.32.164.138]) by core3.amsl.com (Postfix) with ESMTP id 5FB2028C2E5 for <tsvwg@ietf.org>; Fri, 10 Jul 2009 06:11:43 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp3.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 14:12:11 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Fri, 10 Jul 2009 14:12:10 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 124723152960; Fri, 10 Jul 2009 14:12:09 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6ADC48v006312; Fri, 10 Jul 2009 14:12:04 +0100
Message-Id: <200907101312.n6ADC48v006312@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Thu, 09 Jul 2009 22:20:22 +0100
To: <toby.moncaster@bt.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <AEDCAF87EEC94F49BA92EBDD49854CC70C239DDF@E03MVZ1-UKDY.doma in1.systemhost.net>
References: <AEDCAF87EEC94F49BA92EBDD49854CC70C239DDF@E03MVZ1-UKDY.domain1.systemhost.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 10 Jul 2009 13:12:10.0573 (UTC) FILETIME=[08919FD0:01CA0160]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 13:11:49 -0000

Toby,

Thanks for this.
If I say nothing I agree. Otherwise I push back inline...

At 14:42 09/07/2009, toby.moncaster@bt.com wrote:
>Abstract:
>
>Seems to lack detail of what you are expecting people to do having read
>this. Need to clarify that whilst this updates RFC4301 & RFC3168 it only
>ADDS functionality to 4301 (in other words 4301 compliant endpoints will
>always be a subset of those compliant with this standard).

OK. It will have "Updates RFC3168 & RFC4301" in the headers. But as 
so many people have asked for this in the abstract too, I guess I'll 
have to oblige.


>Introduction:
>
>Overall - seems a bit over-long and gives too much detail both about
>WHAT you are intending to do and WHY. I would suggest summarise the what
>and concentrate on the why (as that is going to get less explanation in
>the rest of the document).
>
>Make it clear from the start that you are trying to impose a common set
>of rules on this. Something like "currently the behaviour of IP-in-IP
>tunnel endpoints is not consistent, with consequent risk of
>misconfigurations etc etc..."
>
>Make it more obvious that the changes on entry are to bring all current
>behaviours into line. And thus that the major changes are at the
>egress...
>
>Re-cap of ECN is irrelevant for this document - you can sum this up in 1
>sentence here and move the table into a later section or an appendix...

Agree with the rest of your comments about intro, but not this one. 
This doc has to be read & understood by IPsec implementers and IPinIP 
tunnel implementers who don't have ECN burned in to their skull like we do.


>You still seem to have too much justification text explaining why this
>situation occurred - this could be fine later on but is too detailed for
>the intro.
>
>" As a result, an unfortunate sequence of standards actions has left us
>with nearly the worst of all possible combinations of outcomes, despite
>the best endeavours of everyone concerned."
>Sounds political to me (your telling them they screwed up but you try to
>sugar the pill). Only really need to say that we have ended up with
>contradictions (and I believe it is more that the different groups were
>trying to second-guess each other rather than because they were
>expending their best endeavours).

No, this is actually true as I understand it. When 3168 was written, 
covert channels were identified as a vulnerability (from memory I 
think David Black did in his draft on ECN tunnels at the time). But 
at that time, ECN was a tiny side-show for the security area, so if I 
had been in the security area, I wouldn't have stuck my neck out to 
say covert channels were less important than ECN's control loop. 
Then, when IPsec was being updated, the security area had no right to 
change all tunnels, just because they were changing IPsec tunnels. 
This caused the perversity we have now with only non-IPsec tunnels 
blocking covert channels.

>You mention PCN twice - first as the reason for needing to do something
>now then as the main motivator for doing anything at all. I would
>suggest stick to the first take - PCN highlights the absurdity of the
>current situation so now is the time to do something. However longer
>term this is going to sound strange once PCN is standardised and this is
>accepted so it might be worth pointing out this text is not relevant to
>the standard itself?

I've had other comments asking for substantial reduction of the 
Intro. I'll try.


>1.1 Scope - this seems fine.
>
>1.2 Document Roadmap seems fine (but hides the presence of many more
>appendixes)
>
>Section 3.
>
>Could this become mainly Appendix material? Keep a summary, but cut the
>length?

Michael Menth (I think) suggested I keep only the material 
summarising ECN here, and move material explaining why it needs 
changing to after I specify what to change in Section 4 (when it will 
make more sense). That's what I'll do.

>Are you effectively defining the term "resetting" an ECN
>codepoint? If so have a definitions section...

OK. Thoughts on other terms that could be defined:
Ingress/encap; Egress/decap; Incoming Hdr; Outer Hdr; Inner Hdr; 
Outgoing Hdr; Resetting ECN; Copying ECN.

>"But if non-IPsec tunnels are already present within a network to which
>PCN is being added, that is not particularly helpful advice." - indeed,
>but it is exactly as unhelpful to define a new set of standards
>behaviours!!!

Not so.
- If a standard is defined, an operator can tell a vendor it wants 
equipment that complies with the standard.
- If a standard isn't defined, all an operator can do is read a 
sentence saying "If you want PCN and tunnels: tough - it won't work".


>Section 3.2 can certainly be shorter...
>
>Section 4 is the main point of the document... Make it clear this is the
>case
>
>In figure 4 include a footer describing the meaning of (!!!) as some
>people will fial to read through the actual text...

No need to duplicate. If people want to know what it means, they will 
find where it says what it means. If they don't, they won't.


>4.1
>
>"Compatibility mode is the same per packet behaviour as the ingress end
>of RFC3168's limited functionality mode." --> Compatibility mode behaves
>the same as the limited functionality mode in RFC3168 but only applies
>at the tunnel ingress.

Yes, with the addition of "per-packet", which is important, as a 
tunnel end-point also has set-up behaviour.


>4.2
>First bullet below figure 4 "If the inner ECN field is Not-ECT the
>decapsulator MUST NOT propagate any other ECN codepoint in the outer
>header onwards." --> If the inner ECN field is Not-ECT the decapsulator
>MUST NOT propagate any other ECN codepoint forwards.
>
>First sub-bullet below figure 4 - I believe this may be actually defined
>in RFC4774 as the only safe behaviour. Probably worth spelling out that
>the issue is that the congestion mark may not get reported back to the
>sender.
>
>3rd bullet - would be nice to explain the SHOULD (do you log every time,
>or just log that it has happened; what circs is it ok not to log these)
>
>As section 4.3 is informative it would be clearer to move it into its
>own section (so all the normative text is in a single place).

But, even if it were a section, rather than a sub-section, it would 
be logical to have it at the same point (as it repeats the same 
reasoning for the changes just made in S4.1 & 4.2). So it makes no 
odds whether it's subsection or section, it still goes normative, 
informative, normative.


>End of 1st proper para: "Therefore here we give guidance for designing
>possibly different marking schemes." Is very clunky... do you mean
>guidance for designing different marking schemes, possible different
>marking schemes, marking schemes that possibly differ,...
>
>First bullet: xref to RFC4774 which has strong parallels with this
>advice. I know 4774 is badly written but people are going to notice that
>there is overlap between these and so it would be good to reference it
>(e.g. "RFC4774 provides guidance on overloading the ECN field...")
>
>3rd bullet - the only instance this bullet actually covers is that of
>not ECT on the inner header so why not just say this?
>
>5. End of 1st para. Highlight that this is a big advantage as it is
>likely to reduce possible misconfigurations

I'm not a great fan of repeated self-praise in a standards track doc. 
That can be done on mailing lists when people argue. But if the 
arguments are won, there's no need to have on record how good the 
author thinks their idea is. But I will collect this sort of thing at 
the start where the motivation is written.

>Near the end you have some normative text - this is adrift of the main
>block of such text so should actually be in section 4 (although the
>explanations for it can be here).

I hadn't intended S.5 to be non-normative. But I like your idea. I'll have
- a sub-section of 4 where I give the normative stuff about modes.
- a section 5 that starts with a sub-section "Changes From Earlier 
RFCs" (currently S.6)
- Then further subsections explaining the non-impact of the changes 
on interworking etc.



>Section 6: can this be presented in concise bullets at the start of the
>section? E.g. this updates 3168 in following ways... etc

Gorry has asked that I categorise changes for each RFC separately.


>Section 8 - I suggest having the details of the Nonce hole put into an
>Appendix with an xref from here...
>
>Update references especially RFC5559 PCN!
>
>Appendixes - I take it eventually these will get significantly reduced
>and are really there for the purposes of explanation during
>standardisation? The following comments are if you are going to keep the
>appendixes...
>
>App A.1 This seems to be a mixture of security 101 (as the Americans
>would put it) and specific stuff for this actual work. Do you need all
>the tutorial stuff?

Many in the transport area don't know security 101 (same reason for 
the ECN 101 in the intro for security area people).


>A.2 Sure this could be a bit more concise!

I'll mark it for editing down, but it's not my priority at the mo, 
unless I have time.


>A.3 This seems the right sort of level of detail
>
>B.1 can any of this be explained by a suitable reference?

I'm going to remove the whole of B.

To answer your question, no, the identifier stuff is all new. And the 
rest is all stuff I worked out (using David Black's similar ideas in 
RFC2983) in order to make sense of the interaction between tunnelling 
& sub-path control loops.
It's actually important architectural stuff, but having used it to 
work out what standards actions are appropriate, it's no longer 
necessary in this doc.


>B.2 & B.3 Seem excessively detailed given you now have more or less
>convinced people this is worthwhile work.
>
>Appendix D title is slightly ambiguous (you have to work out where the
>emphasis goes...).

OK



>HTH

Greatly. Thanks.


Bob


>Toby

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From toby.moncaster@bt.com  Fri Jul 10 07:44:17 2009
Return-Path: <toby.moncaster@bt.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 837DF3A6B02 for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 07:44:17 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.644
X-Spam-Level: 
X-Spam-Status: No, score=-2.644 tagged_above=-999 required=5 tests=[AWL=-0.246, BAYES_00=-2.599, J_CHICKENPOX_72=0.6, J_CHICKENPOX_91=0.6, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id OxkfB8ptazmt for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 07:44:16 -0700 (PDT)
Received: from smtp4.smtp.bt.com (smtp4.smtp.bt.com [217.32.164.151]) by core3.amsl.com (Postfix) with ESMTP id 6A51128C33E for <tsvwg@ietf.org>; Fri, 10 Jul 2009 07:44:08 -0700 (PDT)
Received: from E03MVZ1-UKDY.domain1.systemhost.net ([193.113.30.63]) by smtp4.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 15:44:35 +0100
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Fri, 10 Jul 2009 15:43:28 +0100
Message-ID: <AEDCAF87EEC94F49BA92EBDD49854CC70C2C488A@E03MVZ1-UKDY.domain1.systemhost.net>
In-Reply-To: <200907101312.n6ADC48v006312@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
Thread-Index: AcoBYAhGFxyBPOs4T+Sug0Jjg+7hkAAC54JQ
References: <AEDCAF87EEC94F49BA92EBDD49854CC70C239DDF@E03MVZ1-UKDY.domain1.systemhost.net> <200907101312.n6ADC48v006312@bagheera.jungle.bt.co.uk>
From: <toby.moncaster@bt.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 10 Jul 2009 14:44:35.0826 (UTC) FILETIME=[F1CC1520:01CA016C]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 14:44:17 -0000

Couple of responses inline

Toby

> -----Original Message-----
> From: Briscoe,RJ,Bob,XVR9 BRISCORJ R
> Sent: 09 July 2009 22:20
> To: Moncaster,T,Toby,DER3 R
> Cc: tsvwg@ietf.org
> Subject: Re: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
>=20
> Toby,
>=20
> Thanks for this.
> If I say nothing I agree. Otherwise I push back inline...
>=20
> At 14:42 09/07/2009, toby.moncaster@bt.com wrote:
> >Abstract:
> >
> >Seems to lack detail of what you are expecting people to do having
> read
> >this. Need to clarify that whilst this updates RFC4301 & RFC3168 it
> only
> >ADDS functionality to 4301 (in other words 4301 compliant endpoints
> will
> >always be a subset of those compliant with this standard).
>=20
> OK. It will have "Updates RFC3168 & RFC4301" in the headers. But as
> so many people have asked for this in the abstract too, I guess I'll
> have to oblige.
>=20
>=20
> >Introduction:
> >
> >Overall - seems a bit over-long and gives too much detail both about
> >WHAT you are intending to do and WHY. I would suggest summarise the
> what
> >and concentrate on the why (as that is going to get less explanation
> in
> >the rest of the document).
> >
> >Make it clear from the start that you are trying to impose a common
> set
> >of rules on this. Something like "currently the behaviour of IP-in-IP
> >tunnel endpoints is not consistent, with consequent risk of
> >misconfigurations etc etc..."
> >
> >Make it more obvious that the changes on entry are to bring all
> current
> >behaviours into line. And thus that the major changes are at the
> >egress...
> >
> >Re-cap of ECN is irrelevant for this document - you can sum this up
in
> 1
> >sentence here and move the table into a later section or an
> appendix...
>=20
> Agree with the rest of your comments about intro, but not this one.
> This doc has to be read & understood by IPsec implementers and IPinIP
> tunnel implementers who don't have ECN burned in to their skull like
we
> do.

But they don't necessarily need it in detail in the intro was more my
point. Perhaps make it a subsection with a first sentence something like
"recapping ECN for those that are unfamiliar"

>=20
>=20
> >You still seem to have too much justification text explaining why
this
> >situation occurred - this could be fine later on but is too detailed
> for
> >the intro.
> >
> >" As a result, an unfortunate sequence of standards actions has left
> us
> >with nearly the worst of all possible combinations of outcomes,
> despite
> >the best endeavours of everyone concerned."
> >Sounds political to me (your telling them they screwed up but you try
> to
> >sugar the pill). Only really need to say that we have ended up with
> >contradictions (and I believe it is more that the different groups
> were
> >trying to second-guess each other rather than because they were
> >expending their best endeavours).
>=20
> No, this is actually true as I understand it. When 3168 was written,
> covert channels were identified as a vulnerability (from memory I
> think David Black did in his draft on ECN tunnels at the time). But
> at that time, ECN was a tiny side-show for the security area, so if I
> had been in the security area, I wouldn't have stuck my neck out to
> say covert channels were less important than ECN's control loop.
> Then, when IPsec was being updated, the security area had no right to
> change all tunnels, just because they were changing IPsec tunnels.
> This caused the perversity we have now with only non-IPsec tunnels
> blocking covert channels.
>=20
> >You mention PCN twice - first as the reason for needing to do
> something
> >now then as the main motivator for doing anything at all. I would
> >suggest stick to the first take - PCN highlights the absurdity of the
> >current situation so now is the time to do something. However longer
> >term this is going to sound strange once PCN is standardised and this
> is
> >accepted so it might be worth pointing out this text is not relevant
> to
> >the standard itself?
>=20
> I've had other comments asking for substantial reduction of the
> Intro. I'll try.
>=20
>=20
> >1.1 Scope - this seems fine.
> >
> >1.2 Document Roadmap seems fine (but hides the presence of many more
> >appendixes)
> >
> >Section 3.
> >
> >Could this become mainly Appendix material? Keep a summary, but cut
> the
> >length?
>=20
> Michael Menth (I think) suggested I keep only the material
> summarising ECN here, and move material explaining why it needs
> changing to after I specify what to change in Section 4 (when it will
> make more sense). That's what I'll do.
>=20
> >Are you effectively defining the term "resetting" an ECN
> >codepoint? If so have a definitions section...
>=20
> OK. Thoughts on other terms that could be defined:
> Ingress/encap; Egress/decap; Incoming Hdr; Outer Hdr; Inner Hdr;
> Outgoing Hdr; Resetting ECN; Copying ECN.
>=20

Agree - makes exact scope of doc clearer if you define exactly what you
mean by terminology

> >"But if non-IPsec tunnels are already present within a network to
> which
> >PCN is being added, that is not particularly helpful advice." -
> indeed,
> >but it is exactly as unhelpful to define a new set of standards
> >behaviours!!!
>=20
> Not so.
> - If a standard is defined, an operator can tell a vendor it wants
> equipment that complies with the standard.
> - If a standard isn't defined, all an operator can do is read a
> sentence saying "If you want PCN and tunnels: tough - it won't work".
>=20
>=20
> >Section 3.2 can certainly be shorter...
> >
> >Section 4 is the main point of the document... Make it clear this is
> the
> >case
> >
> >In figure 4 include a footer describing the meaning of (!!!) as some
> >people will fial to read through the actual text...
>=20
> No need to duplicate. If people want to know what it means, they will
> find where it says what it means. If they don't, they won't.
>=20

Can't see why it would hurt to clarify in the diagram, but I guess it's
a pretty minor point

>=20
> >4.1
> >
> >"Compatibility mode is the same per packet behaviour as the ingress
> end
> >of RFC3168's limited functionality mode." --> Compatibility mode
> behaves
> >the same as the limited functionality mode in RFC3168 but only
applies
> >at the tunnel ingress.
>=20
> Yes, with the addition of "per-packet", which is important, as a
> tunnel end-point also has set-up behaviour.
>=20
>=20
> >4.2
> >First bullet below figure 4 "If the inner ECN field is Not-ECT the
> >decapsulator MUST NOT propagate any other ECN codepoint in the outer
> >header onwards." --> If the inner ECN field is Not-ECT the
> decapsulator
> >MUST NOT propagate any other ECN codepoint forwards.
> >
> >First sub-bullet below figure 4 - I believe this may be actually
> defined
> >in RFC4774 as the only safe behaviour. Probably worth spelling out
> that
> >the issue is that the congestion mark may not get reported back to
the
> >sender.
> >
> >3rd bullet - would be nice to explain the SHOULD (do you log every
> time,
> >or just log that it has happened; what circs is it ok not to log
> these)
> >
> >As section 4.3 is informative it would be clearer to move it into its
> >own section (so all the normative text is in a single place).
>=20
> But, even if it were a section, rather than a sub-section, it would
> be logical to have it at the same point (as it repeats the same
> reasoning for the changes just made in S4.1 & 4.2). So it makes no
> odds whether it's subsection or section, it still goes normative,
> informative, normative.
>=20
>=20
> >End of 1st proper para: "Therefore here we give guidance for
designing
> >possibly different marking schemes." Is very clunky... do you mean
> >guidance for designing different marking schemes, possible different
> >marking schemes, marking schemes that possibly differ,...
> >
> >First bullet: xref to RFC4774 which has strong parallels with this
> >advice. I know 4774 is badly written but people are going to notice
> that
> >there is overlap between these and so it would be good to reference
it
> >(e.g. "RFC4774 provides guidance on overloading the ECN field...")
> >
> >3rd bullet - the only instance this bullet actually covers is that of
> >not ECT on the inner header so why not just say this?
> >
> >5. End of 1st para. Highlight that this is a big advantage as it is
> >likely to reduce possible misconfigurations
>=20
> I'm not a great fan of repeated self-praise in a standards track doc.
> That can be done on mailing lists when people argue. But if the
> arguments are won, there's no need to have on record how good the
> author thinks their idea is. But I will collect this sort of thing at
> the start where the motivation is written.
>=20
> >Near the end you have some normative text - this is adrift of the
main
> >block of such text so should actually be in section 4 (although the
> >explanations for it can be here).
>=20
> I hadn't intended S.5 to be non-normative. But I like your idea. I'll
> have
> - a sub-section of 4 where I give the normative stuff about modes.
> - a section 5 that starts with a sub-section "Changes From Earlier
> RFCs" (currently S.6)
> - Then further subsections explaining the non-impact of the changes
> on interworking etc.
>=20
>=20
>=20
> >Section 6: can this be presented in concise bullets at the start of
> the
> >section? E.g. this updates 3168 in following ways... etc
>=20
> Gorry has asked that I categorise changes for each RFC separately.
>=20
>=20
> >Section 8 - I suggest having the details of the Nonce hole put into
an
> >Appendix with an xref from here...
> >
> >Update references especially RFC5559 PCN!
> >
> >Appendixes - I take it eventually these will get significantly
reduced
> >and are really there for the purposes of explanation during
> >standardisation? The following comments are if you are going to keep
> the
> >appendixes...
> >
> >App A.1 This seems to be a mixture of security 101 (as the Americans
> >would put it) and specific stuff for this actual work. Do you need
all
> >the tutorial stuff?
>=20
> Many in the transport area don't know security 101 (same reason for
> the ECN 101 in the intro for security area people).
>=20
>=20
> >A.2 Sure this could be a bit more concise!
>=20
> I'll mark it for editing down, but it's not my priority at the mo,
> unless I have time.
>=20
>=20
> >A.3 This seems the right sort of level of detail
> >
> >B.1 can any of this be explained by a suitable reference?
>=20
> I'm going to remove the whole of B.
>=20
> To answer your question, no, the identifier stuff is all new. And the
> rest is all stuff I worked out (using David Black's similar ideas in
> RFC2983) in order to make sense of the interaction between tunnelling
> & sub-path control loops.
> It's actually important architectural stuff, but having used it to
> work out what standards actions are appropriate, it's no longer
> necessary in this doc.
>=20
>=20
> >B.2 & B.3 Seem excessively detailed given you now have more or less
> >convinced people this is worthwhile work.
> >
> >Appendix D title is slightly ambiguous (you have to work out where
the
> >emphasis goes...).
>=20
> OK
>=20
>=20
>=20
> >HTH
>=20
> Greatly. Thanks.
>=20
>=20
> Bob
>=20
>=20
> >Toby
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research


From toby.moncaster@bt.com  Fri Jul 10 07:47:28 2009
Return-Path: <toby.moncaster@bt.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id EE94528C34B for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 07:47:28 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -4.217
X-Spam-Level: 
X-Spam-Status: No, score=-4.217 tagged_above=-999 required=5 tests=[AWL=1.382,  BAYES_00=-2.599, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id njoYdgIM83BY for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 07:47:27 -0700 (PDT)
Received: from smtp1.smtp.bt.com (smtp1.smtp.bt.com [217.32.164.137]) by core3.amsl.com (Postfix) with ESMTP id 41DBB28C349 for <tsvwg@ietf.org>; Fri, 10 Jul 2009 07:47:26 -0700 (PDT)
Received: from E03MVZ1-UKDY.domain1.systemhost.net ([193.113.30.63]) by smtp1.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 15:47:54 +0100
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Fri, 10 Jul 2009 15:47:47 +0100
Message-ID: <AEDCAF87EEC94F49BA92EBDD49854CC70C2C489F@E03MVZ1-UKDY.domain1.systemhost.net>
In-Reply-To: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
Thread-Index: AcoAyKODkrTXRYJHQkO0cbJfaaVkNgApCsxw
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk>
From: <toby.moncaster@bt.com>
To: <rbriscoe@jungle.bt.co.uk>, <Black_David@emc.com>
X-OriginalArrivalTime: 10 Jul 2009 14:47:54.0930 (UTC) FILETIME=[6878F120:01CA016D]
Cc: gorry@erg.abdn.ac.uk, tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 14:47:29 -0000

Just a brief email to say I generally support most of Bob's reasoning
here. Why put in place artificial blocks on currently unused transitions
that just constrain future standards. This really isn't just relevant to
PCN, it just happens that PCN is (seemingly) the first protocol that has
hit this as an obstruction and so it is as a result of PCN that this
issue has come to the fore...

Clearly once you have RFC4774 on guidelines for re-using the ECN
codepoints you need to assume that novel uses for the ECN codepoints
will come up. To then constrain them because of contradictory, outdated
or restrictive tunnelling rules is just digging a hole and reducing
their chance of success, thus stifling innovation in the transport
layer.=20

Toby

> -----Original Message-----
> From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org] On Behalf
> Of Bob Briscoe
> Sent: 09 July 2009 20:08
> To: Black_David@emc.com
> Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>=20
> David,
>=20
> Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...
>=20
> Your comments get to the nub of the proposed egress protocol changes.
> Great... inline...
>=20
> At 00:43 09/07/2009, Black_David@emc.com wrote:
> >Bob,
> >
> >Sorry for the delay on this: things got entirely too busy around the
> >end of the quarter here.  Having re-read the draft from scratch, I
> >have a number of things to comment on ... I will be in Stockholm,
> >and some of these items are going to need further discussion
> >(private and/or public) in Stockholm.
> >
> >Starting with this table in Figure 4:
> >
> >+------------------+----------------------------------------------+
> >|  Incoming Inner  |             Incoming Outer Header            |
> >|      Header      +---------+------------+------------+----------+
> >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> >+------------------+---------+------------+------------+----------+
> >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)| drop(!!!)|
> >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |   CE     |
> >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |   CE     |
> >|        CE        |      CE |     CE     |     CE(!!!)|   CE     |
> >+------------------+---------+------------+------------+----------+
> >                    |                Outgoing Header               |
> >                    +----------------------------------------------+
> >
> >My initial focus is on what needs to change from RFC 3168/RFC 4301
> >behavior and why.  Given that these two RFCs line up, I think changes
> >need strong motivation.  So, here are the four changes, and my
> thoughts:
> >
> >a) Inner: Not-ECT, Outer: ECT(0)
> >         3168/4301: drop (!!!)
> >         Change: Not-ECT (!!!)
> >         Rationale: "allow future standards actions to use this
> >                 combination"
> >         My view: This ain't broke, why are you fixing it?
>=20
> It is broke in the sense tunnel egresses unnecessarily drop harmless
> packets merely because they carry a "currently unused" (CU)
> combination of codepoints. This is similar to the problem ECN faced
> with firewalls dropping packets unless they were Not-ECT.
>=20
> (a) wasn't originally part of the changes I was proposing. But Anil
> Agawaal convinced me it is worth doing. Reasoning: If we are making a
> change to decap anyway, let's get it as right as possible. If we
> decide not to make any other changes to decap, I wouldn't fight for
> this one alone.
>=20
>=20
> >b) Inner: ECT(0), Outer: ECT(1)
> >         3168/4301: ECT(0)
> >         Change: ECT(1)
> >         Rationale: Treat ECT(1) as more severe congestion than
> ECT(0).
> >         My view: Longer discussion below.
> >
> >c) Inner: ECT(1), Outer: ECT(0)
> >         3168/4301: ECT(1)
> >         Change: ECT(1) (!!!)
> >         Rationale: "no currently legal transition" can produce this
> >combo.
> >         My view: Not a good idea, among other reasons because it's
at
> >                 odds with your "future standards action" explanation
> for
> >                 the first change.  I'd remove this change and
instead
> >deal
> >                 with the possibility of raising an alarm in a
> >PCN-specific
> >                 draft rather than declaring this to be an alarm-able
> >combo
> >                 in general.
> >
> >d) Inner: CE, Outer: ECT(1)
> >         3168/4301: CE
> >         Change: CE (!!!)
> >         Rationale & My view: <same as previous change.
>=20
> c) & d) aren't PCN specific.
>=20
> Reasoning for adding alarms (also said this eariler on list in
> conversation with Gorry):
> If security folks have concerns over possible attacks from "currently
> unused" combinations, I want to give them a way to manage the
> perceived threat without just dropping packets. RFC3168 recommended
> alarms for CU combinations. In these cases, I believe RFC3168 simply
> overlooked that these are also CU combinations.
>=20
>=20
> >So, I would remove a) as not defensible (no good reason for change),
> >and I would push the discussion of adding alarms for c) and d) into
> >a PCN-specific document.
>=20
> c) & d) are not to do with PCN.
>=20
> My motivation is: While making change (b) to enable PCN, lets do
> (a,c,d) to make the IP layer as clean as possible, to ensure easier
> future evolvability.
>=20
> I don't see the issue. It says SHOULD log and MAY alarm. So the
> operator doesn't have to do anything.
>=20
> >The c) alarm in particular is only
> >meaningful if PCN or something like it is in use.
>=20
> No. The combination in c) always was CU, and it's still CU with PCN.
> None of the proposed changes are PCN-specific.
>=20
> Even (b) is intended to also be useful for any 3-level ECN while not
> precluding 2-level ECN and the nonce.
>=20
> NB:  3 level =3D 2 'congested' levels + 1 level not congested, e.g. =
VCP
> or PCN which were both mentioned in RFC4774. This doesn't count the
> not-ECN-capable state, which makes 4 states.
>=20
> >I don't view d)
> >as deserving of an alarm in general, as even though something has
> >happened, no damage has been done - this is in line with the
> >"be liberal in what you accept" principle of protocol design.
>=20
> That's why I'm proposing an optional alarm. Optional alarms *are*
> liberal, but they allow operators a way to address security concerns
> about 'known unknowns' [Rumsfelt], while still allowing new
> behaviours to be introduced.
>=20
>=20
> >That leaves b), which I think is the key desired change, because
> >it makes a third ECN code point usable through tunnels.
>=20
> Yes - exactly.
>=20
> >Much as
> >this might be useful for PCN, it may be too late to do this.  The
> >problem starts from the need to say that this draft updates RFC 4301.
> >The result will be a mix of 4301 and updated-4301 implementations
> >that render the third state difficult to use in practice.  The
> >problem's even worse for non-IPsec tunnels.
>=20
> The idea is that an operator wanting to implement PCN will have a
> spec they can point to in their invitation to tender. At the same
> time this spec both UPDATEs and is backward compatible with RFC3168 &
> RFC4301 because it only changes CU stuff at egress and the change at
> the ingress leads to the same output at the egress. Therefore vendors
> can update 3168 & 4301 tunnels lazily - only when their customers ask
> for the new features (whether for PCN or for something else like VCP).
>=20
> >I also read Appendix D
> >as saying that PCN is going the two-state route, not three.
>=20
> That's not how it should be read - I'll clarify it. The PCN w-g wants
> 3 states, but can't have them while ecn-tunnel is not proposed
> standard. Therefore, it has decided to define a 2-state baseline and
> a number of experimental schemes to work round this tunnel issue
> (plus a comparison doc about them all). It has agreed that eventually
> just one experimental scheme will be chosen for standards track:
> - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> - 3-state (which uses an extra DSCP)
> - psdm (which requires out of band state)
> - sm (single marking gets 3 functions from 2 encoding states but
> admits admission control & termination aren't very precise in all
> cases)
>=20
> ___
> See part 2/2 for responses to your comments on specific parts of the
> test.
>=20
> Thanks again for the review
>=20
>=20
> Bob
>=20
>=20
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research


From rbriscoe@jungle.bt.co.uk  Fri Jul 10 08:39:49 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 0FB1F28C104 for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 08:39:49 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.484
X-Spam-Level: 
X-Spam-Status: No, score=-1.484 tagged_above=-999 required=5 tests=[AWL=-0.567, BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, J_CHICKENPOX_72=0.6, J_CHICKENPOX_91=0.6, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 4KJHLYpvbFIY for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 08:39:43 -0700 (PDT)
Received: from smtp1.smtp.bt.com (smtp1.smtp.bt.com [217.32.164.137]) by core3.amsl.com (Postfix) with ESMTP id 3F9D728C0FF for <tsvwg@ietf.org>; Fri, 10 Jul 2009 08:39:22 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp1.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 16:39:50 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Fri, 10 Jul 2009 16:39:48 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247240387360; Fri, 10 Jul 2009 16:39:47 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6AFdgbe011376; Fri, 10 Jul 2009 16:39:42 +0100
Message-Id: <200907101539.n6AFdgbe011376@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Fri, 10 Jul 2009 16:39:50 +0100
To: <toby.moncaster@bt.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <AEDCAF87EEC94F49BA92EBDD49854CC70C2C488A@E03MVZ1-UKDY.doma in1.systemhost.net>
References: <AEDCAF87EEC94F49BA92EBDD49854CC70C239DDF@E03MVZ1-UKDY.domain1.systemhost.net> <200907101312.n6ADC48v006312@bagheera.jungle.bt.co.uk> <AEDCAF87EEC94F49BA92EBDD49854CC70C2C488A@E03MVZ1-UKDY.domain1.systemhost.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 10 Jul 2009 15:39:48.0133 (UTC) FILETIME=[A8160550:01CA0174]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 15:39:49 -0000

Toby,

I'm thinking of putting the ECN codepoints table in the terminology 
section. But it depends whether I can write an intro that makes sense 
without using precise ECN terminology. I'll try.

Adding "recapping ECN for those that are unfamiliar" makes sense.

Reason for avoiding definition of (!!!) in Fig: It needs to be 
precisely defined, and if I put a brief definition in the Fig, 
there's a danger people will only read the Fig and ignore the precise text.
(I'm a culprit, so I know others might do the same.)


Bob

At 15:43 10/07/2009, toby.moncaster@bt.com wrote:
>Couple of responses inline
>
>Toby
>
> > -----Original Message-----
> > From: Briscoe,RJ,Bob,XVR9 BRISCORJ R
> > Sent: 09 July 2009 22:20
> > To: Moncaster,T,Toby,DER3 R
> > Cc: tsvwg@ietf.org
> > Subject: Re: [tsvwg] comments on draft-ietf-tsvwg-ecn-tunnel
> >
> > Toby,
> >
> > Thanks for this.
> > If I say nothing I agree. Otherwise I push back inline...
> >
> > At 14:42 09/07/2009, toby.moncaster@bt.com wrote:
> > >Abstract:
> > >
> > >Seems to lack detail of what you are expecting people to do having
> > read
> > >this. Need to clarify that whilst this updates RFC4301 & RFC3168 it
> > only
> > >ADDS functionality to 4301 (in other words 4301 compliant endpoints
> > will
> > >always be a subset of those compliant with this standard).
> >
> > OK. It will have "Updates RFC3168 & RFC4301" in the headers. But as
> > so many people have asked for this in the abstract too, I guess I'll
> > have to oblige.
> >
> >
> > >Introduction:
> > >
> > >Overall - seems a bit over-long and gives too much detail both about
> > >WHAT you are intending to do and WHY. I would suggest summarise the
> > what
> > >and concentrate on the why (as that is going to get less explanation
> > in
> > >the rest of the document).
> > >
> > >Make it clear from the start that you are trying to impose a common
> > set
> > >of rules on this. Something like "currently the behaviour of IP-in-IP
> > >tunnel endpoints is not consistent, with consequent risk of
> > >misconfigurations etc etc..."
> > >
> > >Make it more obvious that the changes on entry are to bring all
> > current
> > >behaviours into line. And thus that the major changes are at the
> > >egress...
> > >
> > >Re-cap of ECN is irrelevant for this document - you can sum this up
>in
> > 1
> > >sentence here and move the table into a later section or an
> > appendix...
> >
> > Agree with the rest of your comments about intro, but not this one.
> > This doc has to be read & understood by IPsec implementers and IPinIP
> > tunnel implementers who don't have ECN burned in to their skull like
>we
> > do.
>
>But they don't necessarily need it in detail in the intro was more my
>point. Perhaps make it a subsection with a first sentence something like
>"recapping ECN for those that are unfamiliar"
>
> >
> >
> > >You still seem to have too much justification text explaining why
>this
> > >situation occurred - this could be fine later on but is too detailed
> > for
> > >the intro.
> > >
> > >" As a result, an unfortunate sequence of standards actions has left
> > us
> > >with nearly the worst of all possible combinations of outcomes,
> > despite
> > >the best endeavours of everyone concerned."
> > >Sounds political to me (your telling them they screwed up but you try
> > to
> > >sugar the pill). Only really need to say that we have ended up with
> > >contradictions (and I believe it is more that the different groups
> > were
> > >trying to second-guess each other rather than because they were
> > >expending their best endeavours).
> >
> > No, this is actually true as I understand it. When 3168 was written,
> > covert channels were identified as a vulnerability (from memory I
> > think David Black did in his draft on ECN tunnels at the time). But
> > at that time, ECN was a tiny side-show for the security area, so if I
> > had been in the security area, I wouldn't have stuck my neck out to
> > say covert channels were less important than ECN's control loop.
> > Then, when IPsec was being updated, the security area had no right to
> > change all tunnels, just because they were changing IPsec tunnels.
> > This caused the perversity we have now with only non-IPsec tunnels
> > blocking covert channels.
> >
> > >You mention PCN twice - first as the reason for needing to do
> > something
> > >now then as the main motivator for doing anything at all. I would
> > >suggest stick to the first take - PCN highlights the absurdity of the
> > >current situation so now is the time to do something. However longer
> > >term this is going to sound strange once PCN is standardised and this
> > is
> > >accepted so it might be worth pointing out this text is not relevant
> > to
> > >the standard itself?
> >
> > I've had other comments asking for substantial reduction of the
> > Intro. I'll try.
> >
> >
> > >1.1 Scope - this seems fine.
> > >
> > >1.2 Document Roadmap seems fine (but hides the presence of many more
> > >appendixes)
> > >
> > >Section 3.
> > >
> > >Could this become mainly Appendix material? Keep a summary, but cut
> > the
> > >length?
> >
> > Michael Menth (I think) suggested I keep only the material
> > summarising ECN here, and move material explaining why it needs
> > changing to after I specify what to change in Section 4 (when it will
> > make more sense). That's what I'll do.
> >
> > >Are you effectively defining the term "resetting" an ECN
> > >codepoint? If so have a definitions section...
> >
> > OK. Thoughts on other terms that could be defined:
> > Ingress/encap; Egress/decap; Incoming Hdr; Outer Hdr; Inner Hdr;
> > Outgoing Hdr; Resetting ECN; Copying ECN.
> >
>
>Agree - makes exact scope of doc clearer if you define exactly what you
>mean by terminology
>
> > >"But if non-IPsec tunnels are already present within a network to
> > which
> > >PCN is being added, that is not particularly helpful advice." -
> > indeed,
> > >but it is exactly as unhelpful to define a new set of standards
> > >behaviours!!!
> >
> > Not so.
> > - If a standard is defined, an operator can tell a vendor it wants
> > equipment that complies with the standard.
> > - If a standard isn't defined, all an operator can do is read a
> > sentence saying "If you want PCN and tunnels: tough - it won't work".
> >
> >
> > >Section 3.2 can certainly be shorter...
> > >
> > >Section 4 is the main point of the document... Make it clear this is
> > the
> > >case
> > >
> > >In figure 4 include a footer describing the meaning of (!!!) as some
> > >people will fial to read through the actual text...
> >
> > No need to duplicate. If people want to know what it means, they will
> > find where it says what it means. If they don't, they won't.
> >
>
>Can't see why it would hurt to clarify in the diagram, but I guess it's
>a pretty minor point
>
> >
> > >4.1
> > >
> > >"Compatibility mode is the same per packet behaviour as the ingress
> > end
> > >of RFC3168's limited functionality mode." --> Compatibility mode
> > behaves
> > >the same as the limited functionality mode in RFC3168 but only
>applies
> > >at the tunnel ingress.
> >
> > Yes, with the addition of "per-packet", which is important, as a
> > tunnel end-point also has set-up behaviour.
> >
> >
> > >4.2
> > >First bullet below figure 4 "If the inner ECN field is Not-ECT the
> > >decapsulator MUST NOT propagate any other ECN codepoint in the outer
> > >header onwards." --> If the inner ECN field is Not-ECT the
> > decapsulator
> > >MUST NOT propagate any other ECN codepoint forwards.
> > >
> > >First sub-bullet below figure 4 - I believe this may be actually
> > defined
> > >in RFC4774 as the only safe behaviour. Probably worth spelling out
> > that
> > >the issue is that the congestion mark may not get reported back to
>the
> > >sender.
> > >
> > >3rd bullet - would be nice to explain the SHOULD (do you log every
> > time,
> > >or just log that it has happened; what circs is it ok not to log
> > these)
> > >
> > >As section 4.3 is informative it would be clearer to move it into its
> > >own section (so all the normative text is in a single place).
> >
> > But, even if it were a section, rather than a sub-section, it would
> > be logical to have it at the same point (as it repeats the same
> > reasoning for the changes just made in S4.1 & 4.2). So it makes no
> > odds whether it's subsection or section, it still goes normative,
> > informative, normative.
> >
> >
> > >End of 1st proper para: "Therefore here we give guidance for
>designing
> > >possibly different marking schemes." Is very clunky... do you mean
> > >guidance for designing different marking schemes, possible different
> > >marking schemes, marking schemes that possibly differ,...
> > >
> > >First bullet: xref to RFC4774 which has strong parallels with this
> > >advice. I know 4774 is badly written but people are going to notice
> > that
> > >there is overlap between these and so it would be good to reference
>it
> > >(e.g. "RFC4774 provides guidance on overloading the ECN field...")
> > >
> > >3rd bullet - the only instance this bullet actually covers is that of
> > >not ECT on the inner header so why not just say this?
> > >
> > >5. End of 1st para. Highlight that this is a big advantage as it is
> > >likely to reduce possible misconfigurations
> >
> > I'm not a great fan of repeated self-praise in a standards track doc.
> > That can be done on mailing lists when people argue. But if the
> > arguments are won, there's no need to have on record how good the
> > author thinks their idea is. But I will collect this sort of thing at
> > the start where the motivation is written.
> >
> > >Near the end you have some normative text - this is adrift of the
>main
> > >block of such text so should actually be in section 4 (although the
> > >explanations for it can be here).
> >
> > I hadn't intended S.5 to be non-normative. But I like your idea. I'll
> > have
> > - a sub-section of 4 where I give the normative stuff about modes.
> > - a section 5 that starts with a sub-section "Changes From Earlier
> > RFCs" (currently S.6)
> > - Then further subsections explaining the non-impact of the changes
> > on interworking etc.
> >
> >
> >
> > >Section 6: can this be presented in concise bullets at the start of
> > the
> > >section? E.g. this updates 3168 in following ways... etc
> >
> > Gorry has asked that I categorise changes for each RFC separately.
> >
> >
> > >Section 8 - I suggest having the details of the Nonce hole put into
>an
> > >Appendix with an xref from here...
> > >
> > >Update references especially RFC5559 PCN!
> > >
> > >Appendixes - I take it eventually these will get significantly
>reduced
> > >and are really there for the purposes of explanation during
> > >standardisation? The following comments are if you are going to keep
> > the
> > >appendixes...
> > >
> > >App A.1 This seems to be a mixture of security 101 (as the Americans
> > >would put it) and specific stuff for this actual work. Do you need
>all
> > >the tutorial stuff?
> >
> > Many in the transport area don't know security 101 (same reason for
> > the ECN 101 in the intro for security area people).
> >
> >
> > >A.2 Sure this could be a bit more concise!
> >
> > I'll mark it for editing down, but it's not my priority at the mo,
> > unless I have time.
> >
> >
> > >A.3 This seems the right sort of level of detail
> > >
> > >B.1 can any of this be explained by a suitable reference?
> >
> > I'm going to remove the whole of B.
> >
> > To answer your question, no, the identifier stuff is all new. And the
> > rest is all stuff I worked out (using David Black's similar ideas in
> > RFC2983) in order to make sense of the interaction between tunnelling
> > & sub-path control loops.
> > It's actually important architectural stuff, but having used it to
> > work out what standards actions are appropriate, it's no longer
> > necessary in this doc.
> >
> >
> > >B.2 & B.3 Seem excessively detailed given you now have more or less
> > >convinced people this is worthwhile work.
> > >
> > >Appendix D title is slightly ambiguous (you have to work out where
>the
> > >emphasis goes...).
> >
> > OK
> >
> >
> >
> > >HTH
> >
> > Greatly. Thanks.
> >
> >
> > Bob
> >
> >
> > >Toby
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From Black_David@emc.com  Fri Jul 10 09:09:56 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 9CF3A3A6E5E for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 09:09:56 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.203
X-Spam-Level: 
X-Spam-Status: No, score=-7.203 tagged_above=-999 required=5 tests=[AWL=1.396,  BAYES_00=-2.599, GB_I_INVITATION=-2, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id iSsKyZXlXPgT for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 09:09:55 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id 1130D3A6977 for <tsvwg@ietf.org>; Fri, 10 Jul 2009 09:09:54 -0700 (PDT)
Received: from hop04-l1d11-si03.isus.emc.com (HOP04-L1D11-SI03.isus.emc.com [10.254.111.23]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6AGAM01026973 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 10 Jul 2009 12:10:22 -0400
Received: from mailhub.lss.emc.com (numailhub.lss.emc.com [10.254.144.16]) by hop04-l1d11-si03.isus.emc.com (Tablus Interceptor); Fri, 10 Jul 2009 12:10:10 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2/Switch-3.3.2) with ESMTP id n6AGA8qj012704; Fri, 10 Jul 2009 12:10:09 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 12:10:08 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Fri, 10 Jul 2009 12:09:37 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
Thread-Index: AcoAyKUr4Aub5vSPQjiyS6YJsSBzJgAq4FMQ
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 10 Jul 2009 16:10:08.0686 (UTC) FILETIME=[E5385CE0:01CA0178]
X-EMM-EM: Active
Cc: tsvwg@ietf.org, Black_David@emc.com
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 16:09:56 -0000

Quick summary of the four changes proposed in the draft:

a) Inner: Not-ECT, Outer: ECT(0)
	RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
b) Inner: ECT(0), Outer: ECT(1)
	RFC 3168/4301: ECT(0), Change: ECT(1)
c) Inner: ECT(1), Outer: ECT(0)
	RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
d) Inner: CE, Outer: ECT(1)
	RFC 3168/4301: CE, Change: CE(!!!)

If we were designing from scratch, all of these changes would
be fine things to do.  We're not, and especially after the
discussion of how this impacts RFC 2003 gets added, this
draft will be proposing changes to some of the most widely
deployed mechanisms in the Internet.  To me, that strongly
suggests a conservative design approach - focus on what really
needs to be changed, and resist them temptation to fix other
things that would be "nice to have".  In other words, IMHO
the appropriate criteria for "fixing it" is a very strong
cast that "it's broken".

This is an area of principled disagreement on goals between
Bob and myself that list discussion is unlikely to resolve.
My view of the above four changes is that only b) makes it
over the "really needs to be changed" bar (and even then,
only with a *much* stronger statement of the PCN-based
rationale than is in the current draft, and Bob concurs
with strengthening the rationale).

Turning to the other three changes, a) is clearly described
as "nice to have", enough said.

For c) and d), I don't believe the optional alarms are
justified based on ECN (RFC 3168) alone.  A key observation
is that this draft does not make the < b) ECT(0) [Inner]
to ECT(1) [Outer], forward with ECT(1) > case subject to an
optional alarm.  What this means is that this draft not only
enables three codepoints (ECT(0), ECT(1) and CE), but it
*assumes* that they are ordered in terms of priority,
severity, or something analogous.  That is unwarranted as
a general design assumption, IMHO.

For this, reason, I think discussion of which transitions are
optionally alarm-able in this fashion belongs in the context of
a protocol that actually orders these codepoints, not in a
general document that would serve to "just constrain future
standards," borrowing Toby's words ;-).

I could see discussion of optional alarms in an Appendix that
discusses 3-valued PCN as an example, but not in the main body
of the draft.  In the main body, it would be fine to say that
specific uses of the ECN codepoints may assign meanings that
cause some of these transitions to be appropriate for alarms
(and point to the Appendix for a PCN example), but it would
not be ok to constrain the assignment of those meanings by
making the ECT(0) [Inner] to ECT(1) [Outer] change not
optionally alarm-able.

Thanks,
--David
=20

> -----Original Message-----
> From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]=20
> On Behalf Of Bob Briscoe
> Sent: Thursday, July 09, 2009 3:08 PM
> To: Black, David
> Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>=20
> David,
>=20
> Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...
>=20
> Your comments get to the nub of the proposed egress protocol changes.=20
> Great... inline...
>=20
> At 00:43 09/07/2009, Black_David@emc.com wrote:
> >Bob,
> >
> >Sorry for the delay on this: things got entirely too busy around the
> >end of the quarter here.  Having re-read the draft from scratch, I
> >have a number of things to comment on ... I will be in Stockholm,
> >and some of these items are going to need further discussion
> >(private and/or public) in Stockholm.
> >
> >Starting with this table in Figure 4:
> >
> >+------------------+----------------------------------------------+
> >|  Incoming Inner  |             Incoming Outer Header            |
> >|      Header      +---------+------------+------------+----------+
> >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> >+------------------+---------+------------+------------+----------+
> >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)| drop(!!!)|
> >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |   CE     |
> >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |   CE     |
> >|        CE        |      CE |     CE     |     CE(!!!)|   CE     |
> >+------------------+---------+------------+------------+----------+
> >                    |                Outgoing Header               |
> >                    +----------------------------------------------+
> >
> >My initial focus is on what needs to change from RFC 3168/RFC 4301
> >behavior and why.  Given that these two RFCs line up, I think changes
> >need strong motivation.  So, here are the four changes, and=20
> my thoughts:
> >
> >a) Inner: Not-ECT, Outer: ECT(0)
> >         3168/4301: drop (!!!)
> >         Change: Not-ECT (!!!)
> >         Rationale: "allow future standards actions to use this
> >                 combination"
> >         My view: This ain't broke, why are you fixing it?
>=20
> It is broke in the sense tunnel egresses unnecessarily drop harmless=20
> packets merely because they carry a "currently unused" (CU)=20
> combination of codepoints. This is similar to the problem ECN faced=20
> with firewalls dropping packets unless they were Not-ECT.
>=20
> (a) wasn't originally part of the changes I was proposing. But Anil=20
> Agawaal convinced me it is worth doing. Reasoning: If we are making a=20
> change to decap anyway, let's get it as right as possible. If we=20
> decide not to make any other changes to decap, I wouldn't fight for=20
> this one alone.
>=20
>=20
> >b) Inner: ECT(0), Outer: ECT(1)
> >         3168/4301: ECT(0)
> >         Change: ECT(1)
> >         Rationale: Treat ECT(1) as more severe congestion=20
> than ECT(0).
> >         My view: Longer discussion below.
> >
> >c) Inner: ECT(1), Outer: ECT(0)
> >         3168/4301: ECT(1)
> >         Change: ECT(1) (!!!)
> >         Rationale: "no currently legal transition" can=20
> produce this combo.
> >         My view: Not a good idea, among other reasons=20
> because it's at
> >                 odds with your "future standards action"=20
> explanation for
> >                 the first change.  I'd remove this change=20
> and instead deal
> >                 with the possibility of raising an alarm in=20
> a PCN-specific
> >                 draft rather than declaring this to be an=20
> alarm-able combo
> >                 in general.
> >
> >d) Inner: CE, Outer: ECT(1)
> >         3168/4301: CE
> >         Change: CE (!!!)
> >         Rationale & My view: <same as previous change>
>=20
> c) & d) aren't PCN specific.
>=20
> Reasoning for adding alarms (also said this eariler on list in=20
> conversation with Gorry):
> If security folks have concerns over possible attacks from "currently=20
> unused" combinations, I want to give them a way to manage the=20
> perceived threat without just dropping packets. RFC3168 recommended=20
> alarms for CU combinations. In these cases, I believe RFC3168 simply=20
> overlooked that these are also CU combinations.
>=20
>=20
> >So, I would remove a) as not defensible (no good reason for change),
> >and I would push the discussion of adding alarms for c) and d) into
> >a PCN-specific document.
>=20
> c) & d) are not to do with PCN.
>=20
> My motivation is: While making change (b) to enable PCN, lets do=20
> (a,c,d) to make the IP layer as clean as possible, to ensure easier=20
> future evolvability.
>=20
> I don't see the issue. It says SHOULD log and MAY alarm. So the=20
> operator doesn't have to do anything.
>=20
> >The c) alarm in particular is only
> >meaningful if PCN or something like it is in use.
>=20
> No. The combination in c) always was CU, and it's still CU with PCN.=20
> None of the proposed changes are PCN-specific.
>=20
> Even (b) is intended to also be useful for any 3-level ECN while not=20
> precluding 2-level ECN and the nonce.
>=20
> NB:  3 level =3D 2 'congested' levels + 1 level not congested, e.g. =
VCP=20
> or PCN which were both mentioned in RFC4774. This doesn't count the=20
> not-ECN-capable state, which makes 4 states.
>=20
> >I don't view d)
> >as deserving of an alarm in general, as even though something has
> >happened, no damage has been done - this is in line with the
> >"be liberal in what you accept" principle of protocol design.
>=20
> That's why I'm proposing an optional alarm. Optional alarms *are*=20
> liberal, but they allow operators a way to address security concerns=20
> about 'known unknowns' [Rumsfelt], while still allowing new=20
> behaviours to be introduced.
>=20
>=20
> >That leaves b), which I think is the key desired change, because
> >it makes a third ECN code point usable through tunnels.
>=20
> Yes - exactly.
>=20
> >Much as
> >this might be useful for PCN, it may be too late to do this.  The
> >problem starts from the need to say that this draft updates RFC 4301.
> >The result will be a mix of 4301 and updated-4301 implementations
> >that render the third state difficult to use in practice.  The
> >problem's even worse for non-IPsec tunnels.
>=20
> The idea is that an operator wanting to implement PCN will have a=20
> spec they can point to in their invitation to tender. At the same=20
> time this spec both UPDATEs and is backward compatible with RFC3168 &=20
> RFC4301 because it only changes CU stuff at egress and the change at=20
> the ingress leads to the same output at the egress. Therefore vendors=20
> can update 3168 & 4301 tunnels lazily - only when their customers ask=20
> for the new features (whether for PCN or for something else like VCP).
>=20
> >I also read Appendix D
> >as saying that PCN is going the two-state route, not three.
>=20
> That's not how it should be read - I'll clarify it. The PCN w-g wants=20
> 3 states, but can't have them while ecn-tunnel is not proposed=20
> standard. Therefore, it has decided to define a 2-state baseline and=20
> a number of experimental schemes to work round this tunnel issue=20
> (plus a comparison doc about them all). It has agreed that eventually=20
> just one experimental scheme will be chosen for standards track:
> - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> - 3-state (which uses an extra DSCP)
> - psdm (which requires out of band state)
> - sm (single marking gets 3 functions from 2 encoding states but=20
> admits admission control & termination aren't very precise in=20
> all cases)
>=20
> ___
> See part 2/2 for responses to your comments on specific parts=20
> of the test.
>=20
> Thanks again for the review
>=20
>=20
> Bob
>=20
>=20
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research =20
>=20
>=20
>=20

From rbriscoe@jungle.bt.co.uk  Fri Jul 10 10:31:25 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E981E28C104 for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 10:31:25 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.037
X-Spam-Level: 
X-Spam-Status: No, score=-3.037 tagged_above=-999 required=5 tests=[AWL=1.080,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id QhXUois1d7tV for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 10:31:19 -0700 (PDT)
Received: from smtp3.smtp.bt.com (smtp3.smtp.bt.com [217.32.164.138]) by core3.amsl.com (Postfix) with ESMTP id B71BD3A67AD for <tsvwg@ietf.org>; Fri, 10 Jul 2009 10:31:18 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp3.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 18:31:46 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Fri, 10 Jul 2009 18:31:46 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247247105833; Fri, 10 Jul 2009 18:31:45 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6AHVfRf015097; Fri, 10 Jul 2009 18:31:41 +0100
Message-Id: <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Fri, 10 Jul 2009 18:31:37 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 10 Jul 2009 17:31:46.0188 (UTC) FILETIME=[4C5BDCC0:01CA0184]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 17:31:26 -0000

David,

I'm not so pessimistic about reaching consensus on the list. This is 
all about the 3 "nice to have" things (a,c,d). As I've already said, 
none are anything to do with PCN. All the controversial combinations 
are currently unused by ECN and currently unused by PCN.

Our disagreement is over this abstract principle:
When changing something that's needed, is it worse or better to leave 
other things unchanged that would be "nice to have" (ie. things that 
are closing off  future evolvability)?

You say it's better to leave them unchanged. Why?

I say it's worse to leave them unchanged, because you get a big 
potential future gain for minimal extra effort. It costs diddly-squat 
to make 4 changes to a code block if you're already going to make 1 
change. It requires nearly the same testing, etc etc. In return you 
get a greater chance of 3 extra usable codepoint combinations in the IP header.

I look at the 5 different ways that ECN tunnelling has gone through 
in the past, and I think "if we're going to add a 6th, let's avoid a 7th".



Bob

At 17:09 10/07/2009, Black_David@emc.com wrote:
>Quick summary of the four changes proposed in the draft:
>
>a) Inner: Not-ECT, Outer: ECT(0)
>         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
>b) Inner: ECT(0), Outer: ECT(1)
>         RFC 3168/4301: ECT(0), Change: ECT(1)
>c) Inner: ECT(1), Outer: ECT(0)
>         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
>d) Inner: CE, Outer: ECT(1)
>         RFC 3168/4301: CE, Change: CE(!!!)
>
>If we were designing from scratch, all of these changes would
>be fine things to do.  We're not, and especially after the
>discussion of how this impacts RFC 2003 gets added, this
>draft will be proposing changes to some of the most widely
>deployed mechanisms in the Internet.  To me, that strongly
>suggests a conservative design approach - focus on what really
>needs to be changed, and resist them temptation to fix other
>things that would be "nice to have".  In other words, IMHO
>the appropriate criteria for "fixing it" is a very strong
>cast that "it's broken".
>
>This is an area of principled disagreement on goals between
>Bob and myself that list discussion is unlikely to resolve.
>My view of the above four changes is that only b) makes it
>over the "really needs to be changed" bar (and even then,
>only with a *much* stronger statement of the PCN-based
>rationale than is in the current draft, and Bob concurs
>with strengthening the rationale).
>
>Turning to the other three changes, a) is clearly described
>as "nice to have", enough said.
>
>For c) and d), I don't believe the optional alarms are
>justified based on ECN (RFC 3168) alone.  A key observation
>is that this draft does not make the < b) ECT(0) [Inner]
>to ECT(1) [Outer], forward with ECT(1) > case subject to an
>optional alarm.  What this means is that this draft not only
>enables three codepoints (ECT(0), ECT(1) and CE), but it
>*assumes* that they are ordered in terms of priority,
>severity, or something analogous.  That is unwarranted as
>a general design assumption, IMHO.
>
>For this, reason, I think discussion of which transitions are
>optionally alarm-able in this fashion belongs in the context of
>a protocol that actually orders these codepoints, not in a
>general document that would serve to "just constrain future
>standards," borrowing Toby's words ;-).
>
>I could see discussion of optional alarms in an Appendix that
>discusses 3-valued PCN as an example, but not in the main body
>of the draft.  In the main body, it would be fine to say that
>specific uses of the ECN codepoints may assign meanings that
>cause some of these transitions to be appropriate for alarms
>(and point to the Appendix for a PCN example), but it would
>not be ok to constrain the assignment of those meanings by
>making the ECT(0) [Inner] to ECT(1) [Outer] change not
>optionally alarm-able.
>
>Thanks,
>--David
>
>
> > -----Original Message-----
> > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > On Behalf Of Bob Briscoe
> > Sent: Thursday, July 09, 2009 3:08 PM
> > To: Black, David
> > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> >
> > David,
> >
> > Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...
> >
> > Your comments get to the nub of the proposed egress protocol changes.
> > Great... inline...
> >
> > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > >Bob,
> > >
> > >Sorry for the delay on this: things got entirely too busy around the
> > >end of the quarter here.  Having re-read the draft from scratch, I
> > >have a number of things to comment on ... I will be in Stockholm,
> > >and some of these items are going to need further discussion
> > >(private and/or public) in Stockholm.
> > >
> > >Starting with this table in Figure 4:
> > >
> > >+------------------+----------------------------------------------+
> > >|  Incoming Inner  |             Incoming Outer Header            |
> > >|      Header      +---------+------------+------------+----------+
> > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> > >+------------------+---------+------------+------------+----------+
> > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)| drop(!!!)|
> > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |   CE     |
> > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |   CE     |
> > >|        CE        |      CE |     CE     |     CE(!!!)|   CE     |
> > >+------------------+---------+------------+------------+----------+
> > >                    |                Outgoing Header               |
> > >                    +----------------------------------------------+
> > >
> > >My initial focus is on what needs to change from RFC 3168/RFC 4301
> > >behavior and why.  Given that these two RFCs line up, I think changes
> > >need strong motivation.  So, here are the four changes, and
> > my thoughts:
> > >
> > >a) Inner: Not-ECT, Outer: ECT(0)
> > >         3168/4301: drop (!!!)
> > >         Change: Not-ECT (!!!)
> > >         Rationale: "allow future standards actions to use this
> > >                 combination"
> > >         My view: This ain't broke, why are you fixing it?
> >
> > It is broke in the sense tunnel egresses unnecessarily drop harmless
> > packets merely because they carry a "currently unused" (CU)
> > combination of codepoints. This is similar to the problem ECN faced
> > with firewalls dropping packets unless they were Not-ECT.
> >
> > (a) wasn't originally part of the changes I was proposing. But Anil
> > Agawaal convinced me it is worth doing. Reasoning: If we are making a
> > change to decap anyway, let's get it as right as possible. If we
> > decide not to make any other changes to decap, I wouldn't fight for
> > this one alone.
> >
> >
> > >b) Inner: ECT(0), Outer: ECT(1)
> > >         3168/4301: ECT(0)
> > >         Change: ECT(1)
> > >         Rationale: Treat ECT(1) as more severe congestion
> > than ECT(0).
> > >         My view: Longer discussion below.
> > >
> > >c) Inner: ECT(1), Outer: ECT(0)
> > >         3168/4301: ECT(1)
> > >         Change: ECT(1) (!!!)
> > >         Rationale: "no currently legal transition" can
> > produce this combo.
> > >         My view: Not a good idea, among other reasons
> > because it's at
> > >                 odds with your "future standards action"
> > explanation for
> > >                 the first change.  I'd remove this change
> > and instead deal
> > >                 with the possibility of raising an alarm in
> > a PCN-specific
> > >                 draft rather than declaring this to be an
> > alarm-able combo
> > >                 in general.
> > >
> > >d) Inner: CE, Outer: ECT(1)
> > >         3168/4301: CE
> > >         Change: CE (!!!)
> > >         Rationale & My view: <same as previous change>
> >
> > c) & d) aren't PCN specific.
> >
> > Reasoning for adding alarms (also said this eariler on list in
> > conversation with Gorry):
> > If security folks have concerns over possible attacks from "currently
> > unused" combinations, I want to give them a way to manage the
> > perceived threat without just dropping packets. RFC3168 recommended
> > alarms for CU combinations. In these cases, I believe RFC3168 simply
> > overlooked that these are also CU combinations.
> >
> >
> > >So, I would remove a) as not defensible (no good reason for change),
> > >and I would push the discussion of adding alarms for c) and d) into
> > >a PCN-specific document.
> >
> > c) & d) are not to do with PCN.
> >
> > My motivation is: While making change (b) to enable PCN, lets do
> > (a,c,d) to make the IP layer as clean as possible, to ensure easier
> > future evolvability.
> >
> > I don't see the issue. It says SHOULD log and MAY alarm. So the
> > operator doesn't have to do anything.
> >
> > >The c) alarm in particular is only
> > >meaningful if PCN or something like it is in use.
> >
> > No. The combination in c) always was CU, and it's still CU with PCN.
> > None of the proposed changes are PCN-specific.
> >
> > Even (b) is intended to also be useful for any 3-level ECN while not
> > precluding 2-level ECN and the nonce.
> >
> > NB:  3 level = 2 'congested' levels + 1 level not congested, e.g. VCP
> > or PCN which were both mentioned in RFC4774. This doesn't count the
> > not-ECN-capable state, which makes 4 states.
> >
> > >I don't view d)
> > >as deserving of an alarm in general, as even though something has
> > >happened, no damage has been done - this is in line with the
> > >"be liberal in what you accept" principle of protocol design.
> >
> > That's why I'm proposing an optional alarm. Optional alarms *are*
> > liberal, but they allow operators a way to address security concerns
> > about 'known unknowns' [Rumsfelt], while still allowing new
> > behaviours to be introduced.
> >
> >
> > >That leaves b), which I think is the key desired change, because
> > >it makes a third ECN code point usable through tunnels.
> >
> > Yes - exactly.
> >
> > >Much as
> > >this might be useful for PCN, it may be too late to do this.  The
> > >problem starts from the need to say that this draft updates RFC 4301.
> > >The result will be a mix of 4301 and updated-4301 implementations
> > >that render the third state difficult to use in practice.  The
> > >problem's even worse for non-IPsec tunnels.
> >
> > The idea is that an operator wanting to implement PCN will have a
> > spec they can point to in their invitation to tender. At the same
> > time this spec both UPDATEs and is backward compatible with RFC3168 &
> > RFC4301 because it only changes CU stuff at egress and the change at
> > the ingress leads to the same output at the egress. Therefore vendors
> > can update 3168 & 4301 tunnels lazily - only when their customers ask
> > for the new features (whether for PCN or for something else like VCP).
> >
> > >I also read Appendix D
> > >as saying that PCN is going the two-state route, not three.
> >
> > That's not how it should be read - I'll clarify it. The PCN w-g wants
> > 3 states, but can't have them while ecn-tunnel is not proposed
> > standard. Therefore, it has decided to define a 2-state baseline and
> > a number of experimental schemes to work round this tunnel issue
> > (plus a comparison doc about them all). It has agreed that eventually
> > just one experimental scheme will be chosen for standards track:
> > - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> > - 3-state (which uses an extra DSCP)
> > - psdm (which requires out of band state)
> > - sm (single marking gets 3 functions from 2 encoding states but
> > admits admission control & termination aren't very precise in
> > all cases)
> >
> > ___
> > See part 2/2 for responses to your comments on specific parts
> > of the test.
> >
> > Thanks again for the review
> >
> >
> > Bob
> >
> >
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From Black_David@emc.com  Fri Jul 10 15:29:21 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 8539F28C3B7 for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 15:29:21 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.31
X-Spam-Level: 
X-Spam-Status: No, score=-7.31 tagged_above=-999 required=5 tests=[AWL=1.289,  BAYES_00=-2.599, GB_I_INVITATION=-2, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id UOkdpvcsd9GZ for <tsvwg@core3.amsl.com>; Fri, 10 Jul 2009 15:29:19 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id ECBA33A6976 for <tsvwg@ietf.org>; Fri, 10 Jul 2009 15:28:47 -0700 (PDT)
Received: from hop04-l1d11-si04.isus.emc.com (HOP04-L1D11-SI04.isus.emc.com [10.254.111.24]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6AMTFvf027255 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 10 Jul 2009 18:29:15 -0400
Received: from mailhub.lss.emc.com (nagas.lss.emc.com [10.254.144.15]) by hop04-l1d11-si04.isus.emc.com (Tablus Interceptor); Fri, 10 Jul 2009 18:29:05 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2mp/Switch-3.3.2mp) with ESMTP id n6AMSp1U011486; Fri, 10 Jul 2009 18:28:52 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 10 Jul 2009 18:28:52 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Fri, 10 Jul 2009 18:28:23 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
Thread-Index: AcoBhKEVDMVWzytlSIC7PVGxDxy6lQAKAmPg
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 10 Jul 2009 22:28:52.0627 (UTC) FILETIME=[CDBE9630:01CA01AD]
X-EMM-EM: Active
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 10 Jul 2009 22:29:21 -0000

Bob,

Actually it's only about a) if you can accept the following
approach for the optional alarms in c) and d):

- Main body of draft says that specific uses of ECN may raise
	optional decapsulation alarms for selected cases in which
	the inner and outer header values don't match.
- An Appendix that discusses PCN as an example points out that
	for PCN, it is appropriate to raise alarms for c) and d).

As for "How does this have anything to do with PCN?", the answer
to that question is the same as the answer to "Why does the
draft not propose an optional alarm for case b)?" ...  the
answer is probably some form of "no current use", in which case,
what's the current use for the c) and d) alarms? ... it's PCN.

Thanks,
--David
=20

> -----Original Message-----
> From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]=20
> Sent: Friday, July 10, 2009 1:32 PM
> To: Black, David
> Cc: tsvwg@ietf.org
> Subject: RE: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>=20
> David,
>=20
> I'm not so pessimistic about reaching consensus on the list. This is=20
> all about the 3 "nice to have" things (a,c,d). As I've already said,=20
> none are anything to do with PCN. All the controversial combinations=20
> are currently unused by ECN and currently unused by PCN.
>=20
> Our disagreement is over this abstract principle:
> When changing something that's needed, is it worse or better to leave=20
> other things unchanged that would be "nice to have" (ie. things that=20
> are closing off  future evolvability)?
>=20
> You say it's better to leave them unchanged. Why?
>=20
> I say it's worse to leave them unchanged, because you get a big=20
> potential future gain for minimal extra effort. It costs diddly-squat=20
> to make 4 changes to a code block if you're already going to make 1=20
> change. It requires nearly the same testing, etc etc. In return you=20
> get a greater chance of 3 extra usable codepoint combinations=20
> in the IP header.
>=20
> I look at the 5 different ways that ECN tunnelling has gone through=20
> in the past, and I think "if we're going to add a 6th, let's=20
> avoid a 7th".
>=20
>=20
>=20
> Bob
>=20
> At 17:09 10/07/2009, Black_David@emc.com wrote:
> >Quick summary of the four changes proposed in the draft:
> >
> >a) Inner: Not-ECT, Outer: ECT(0)
> >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> >b) Inner: ECT(0), Outer: ECT(1)
> >         RFC 3168/4301: ECT(0), Change: ECT(1)
> >c) Inner: ECT(1), Outer: ECT(0)
> >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> >d) Inner: CE, Outer: ECT(1)
> >         RFC 3168/4301: CE, Change: CE(!!!)
> >
> >If we were designing from scratch, all of these changes would
> >be fine things to do.  We're not, and especially after the
> >discussion of how this impacts RFC 2003 gets added, this
> >draft will be proposing changes to some of the most widely
> >deployed mechanisms in the Internet.  To me, that strongly
> >suggests a conservative design approach - focus on what really
> >needs to be changed, and resist them temptation to fix other
> >things that would be "nice to have".  In other words, IMHO
> >the appropriate criteria for "fixing it" is a very strong
> >cast that "it's broken".
> >
> >This is an area of principled disagreement on goals between
> >Bob and myself that list discussion is unlikely to resolve.
> >My view of the above four changes is that only b) makes it
> >over the "really needs to be changed" bar (and even then,
> >only with a *much* stronger statement of the PCN-based
> >rationale than is in the current draft, and Bob concurs
> >with strengthening the rationale).
> >
> >Turning to the other three changes, a) is clearly described
> >as "nice to have", enough said.
> >
> >For c) and d), I don't believe the optional alarms are
> >justified based on ECN (RFC 3168) alone.  A key observation
> >is that this draft does not make the < b) ECT(0) [Inner]
> >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> >optional alarm.  What this means is that this draft not only
> >enables three codepoints (ECT(0), ECT(1) and CE), but it
> >*assumes* that they are ordered in terms of priority,
> >severity, or something analogous.  That is unwarranted as
> >a general design assumption, IMHO.
> >
> >For this, reason, I think discussion of which transitions are
> >optionally alarm-able in this fashion belongs in the context of
> >a protocol that actually orders these codepoints, not in a
> >general document that would serve to "just constrain future
> >standards," borrowing Toby's words ;-).
> >
> >I could see discussion of optional alarms in an Appendix that
> >discusses 3-valued PCN as an example, but not in the main body
> >of the draft.  In the main body, it would be fine to say that
> >specific uses of the ECN codepoints may assign meanings that
> >cause some of these transitions to be appropriate for alarms
> >(and point to the Appendix for a PCN example), but it would
> >not be ok to constrain the assignment of those meanings by
> >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> >optionally alarm-able.
> >
> >Thanks,
> >--David
> >
> >
> > > -----Original Message-----
> > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > > On Behalf Of Bob Briscoe
> > > Sent: Thursday, July 09, 2009 3:08 PM
> > > To: Black, David
> > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > > Subject: Re: [tsvwg] review of=20
> draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > >
> > > David,
> > >
> > > Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...
> > >
> > > Your comments get to the nub of the proposed egress=20
> protocol changes.
> > > Great... inline...
> > >
> > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > > >Bob,
> > > >
> > > >Sorry for the delay on this: things got entirely too=20
> busy around the
> > > >end of the quarter here.  Having re-read the draft from=20
> scratch, I
> > > >have a number of things to comment on ... I will be in Stockholm,
> > > >and some of these items are going to need further discussion
> > > >(private and/or public) in Stockholm.
> > > >
> > > >Starting with this table in Figure 4:
> > > >
> > >=20
> >+------------------+----------------------------------------------+
> > > >|  Incoming Inner  |             Incoming Outer Header  =20
>          |
> > > >|      Header     =20
> +---------+------------+------------+----------+
> > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |=20
>    CE    |
> > >=20
> >+------------------+---------+------------+------------+----------+
> > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|=20
> drop(!!!)|
> > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |=20
>   CE     |
> > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |=20
>   CE     |
> > > >|        CE        |      CE |     CE     |     CE(!!!)|=20
>   CE     |
> > >=20
> >+------------------+---------+------------+------------+----------+
> > > >                    |                Outgoing Header    =20
>           |
> > > >                   =20
> +----------------------------------------------+
> > > >
> > > >My initial focus is on what needs to change from RFC=20
> 3168/RFC 4301
> > > >behavior and why.  Given that these two RFCs line up, I=20
> think changes
> > > >need strong motivation.  So, here are the four changes, and
> > > my thoughts:
> > > >
> > > >a) Inner: Not-ECT, Outer: ECT(0)
> > > >         3168/4301: drop (!!!)
> > > >         Change: Not-ECT (!!!)
> > > >         Rationale: "allow future standards actions to use this
> > > >                 combination"
> > > >         My view: This ain't broke, why are you fixing it?
> > >
> > > It is broke in the sense tunnel egresses unnecessarily=20
> drop harmless
> > > packets merely because they carry a "currently unused" (CU)
> > > combination of codepoints. This is similar to the problem=20
> ECN faced
> > > with firewalls dropping packets unless they were Not-ECT.
> > >
> > > (a) wasn't originally part of the changes I was=20
> proposing. But Anil
> > > Agawaal convinced me it is worth doing. Reasoning: If we=20
> are making a
> > > change to decap anyway, let's get it as right as possible. If we
> > > decide not to make any other changes to decap, I wouldn't=20
> fight for
> > > this one alone.
> > >
> > >
> > > >b) Inner: ECT(0), Outer: ECT(1)
> > > >         3168/4301: ECT(0)
> > > >         Change: ECT(1)
> > > >         Rationale: Treat ECT(1) as more severe congestion
> > > than ECT(0).
> > > >         My view: Longer discussion below.
> > > >
> > > >c) Inner: ECT(1), Outer: ECT(0)
> > > >         3168/4301: ECT(1)
> > > >         Change: ECT(1) (!!!)
> > > >         Rationale: "no currently legal transition" can
> > > produce this combo.
> > > >         My view: Not a good idea, among other reasons
> > > because it's at
> > > >                 odds with your "future standards action"
> > > explanation for
> > > >                 the first change.  I'd remove this change
> > > and instead deal
> > > >                 with the possibility of raising an alarm in
> > > a PCN-specific
> > > >                 draft rather than declaring this to be an
> > > alarm-able combo
> > > >                 in general.
> > > >
> > > >d) Inner: CE, Outer: ECT(1)
> > > >         3168/4301: CE
> > > >         Change: CE (!!!)
> > > >         Rationale & My view: <same as previous change>
> > >
> > > c) & d) aren't PCN specific.
> > >
> > > Reasoning for adding alarms (also said this eariler on list in
> > > conversation with Gorry):
> > > If security folks have concerns over possible attacks=20
> from "currently
> > > unused" combinations, I want to give them a way to manage the
> > > perceived threat without just dropping packets. RFC3168=20
> recommended
> > > alarms for CU combinations. In these cases, I believe=20
> RFC3168 simply
> > > overlooked that these are also CU combinations.
> > >
> > >
> > > >So, I would remove a) as not defensible (no good reason=20
> for change),
> > > >and I would push the discussion of adding alarms for c)=20
> and d) into
> > > >a PCN-specific document.
> > >
> > > c) & d) are not to do with PCN.
> > >
> > > My motivation is: While making change (b) to enable PCN, lets do
> > > (a,c,d) to make the IP layer as clean as possible, to=20
> ensure easier
> > > future evolvability.
> > >
> > > I don't see the issue. It says SHOULD log and MAY alarm. So the
> > > operator doesn't have to do anything.
> > >
> > > >The c) alarm in particular is only
> > > >meaningful if PCN or something like it is in use.
> > >
> > > No. The combination in c) always was CU, and it's still=20
> CU with PCN.
> > > None of the proposed changes are PCN-specific.
> > >
> > > Even (b) is intended to also be useful for any 3-level=20
> ECN while not
> > > precluding 2-level ECN and the nonce.
> > >
> > > NB:  3 level =3D 2 'congested' levels + 1 level not=20
> congested, e.g. VCP
> > > or PCN which were both mentioned in RFC4774. This doesn't=20
> count the
> > > not-ECN-capable state, which makes 4 states.
> > >
> > > >I don't view d)
> > > >as deserving of an alarm in general, as even though something has
> > > >happened, no damage has been done - this is in line with the
> > > >"be liberal in what you accept" principle of protocol design.
> > >
> > > That's why I'm proposing an optional alarm. Optional alarms *are*
> > > liberal, but they allow operators a way to address=20
> security concerns
> > > about 'known unknowns' [Rumsfelt], while still allowing new
> > > behaviours to be introduced.
> > >
> > >
> > > >That leaves b), which I think is the key desired change, because
> > > >it makes a third ECN code point usable through tunnels.
> > >
> > > Yes - exactly.
> > >
> > > >Much as
> > > >this might be useful for PCN, it may be too late to do this.  The
> > > >problem starts from the need to say that this draft=20
> updates RFC 4301.
> > > >The result will be a mix of 4301 and updated-4301 implementations
> > > >that render the third state difficult to use in practice.  The
> > > >problem's even worse for non-IPsec tunnels.
> > >
> > > The idea is that an operator wanting to implement PCN will have a
> > > spec they can point to in their invitation to tender. At the same
> > > time this spec both UPDATEs and is backward compatible=20
> with RFC3168 &
> > > RFC4301 because it only changes CU stuff at egress and=20
> the change at
> > > the ingress leads to the same output at the egress.=20
> Therefore vendors
> > > can update 3168 & 4301 tunnels lazily - only when their=20
> customers ask
> > > for the new features (whether for PCN or for something=20
> else like VCP).
> > >
> > > >I also read Appendix D
> > > >as saying that PCN is going the two-state route, not three.
> > >
> > > That's not how it should be read - I'll clarify it. The=20
> PCN w-g wants
> > > 3 states, but can't have them while ecn-tunnel is not proposed
> > > standard. Therefore, it has decided to define a 2-state=20
> baseline and
> > > a number of experimental schemes to work round this tunnel issue
> > > (plus a comparison doc about them all). It has agreed=20
> that eventually
> > > just one experimental scheme will be chosen for standards track:
> > > - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> > > - 3-state (which uses an extra DSCP)
> > > - psdm (which requires out of band state)
> > > - sm (single marking gets 3 functions from 2 encoding states but
> > > admits admission control & termination aren't very precise in
> > > all cases)
> > >
> > > ___
> > > See part 2/2 for responses to your comments on specific parts
> > > of the test.
> > >
> > > Thanks again for the review
> > >
> > >
> > > Bob
> > >
> > >
> > >
> > > ________________________________________________________________
> > > Bob Briscoe,               Networks Research Centre, BT Research
> > >
> > >
> > >
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research=20
>=20
>=20
>=20

From rbriscoe@jungle.bt.co.uk  Mon Jul 13 04:06:48 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 345A428C36A for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 04:06:48 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.12
X-Spam-Level: 
X-Spam-Status: No, score=-3.12 tagged_above=-999 required=5 tests=[AWL=0.997,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id Edr9krVJ4H04 for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 04:06:42 -0700 (PDT)
Received: from smtp4.smtp.bt.com (smtp4.smtp.bt.com [217.32.164.151]) by core3.amsl.com (Postfix) with ESMTP id D5FE828C264 for <tsvwg@ietf.org>; Mon, 13 Jul 2009 04:06:41 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp4.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 13 Jul 2009 12:07:11 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 13 Jul 2009 12:07:10 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247483230639; Mon, 13 Jul 2009 12:07:10 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6DB74cn016300; Mon, 13 Jul 2009 12:07:04 +0100
Message-Id: <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 13 Jul 2009 12:07:09 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 13 Jul 2009 11:07:10.0928 (UTC) FILETIME=[11ABDD00:01CA03AA]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 13 Jul 2009 11:06:48 -0000

David,

We're getting somewhere (if we set aside (a) for now).

I'll first try to reverse engineer how the way we think differs...
- You seem to be thinking of different ECN tunnel behaviours for 
different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
- My goal with this draft is to find a lowest common denominator ECN 
tunnel behaviour for all types of tunnel (IPsec, non-IPsec, 3-level, 
2-level, ECN not supported, whatever).

This separation is important in both directions:
- all nodes should be able to expect the ECN field to behave in the 
same way whether or not tunnels intervene without their control
- all tunnels should be able to process the ECN field in the same 
way, whatever ECN is being used for, because tunnels get deployed 
independently of congestion control schemes like ECN & PCN.

However, at encap I've had to accept that just one mode is 
impossible, as we still need the two RFC3168 behaviours for safety 
with legacy egresses.

now inline...

At 23:28 10/07/2009, Black_David@emc.com wrote:
>Bob,
>
>Actually it's only about a) if you can accept the following
>approach for the optional alarms in c) and d):
>
>- Main body of draft says that specific uses of ECN may raise
>         optional decapsulation alarms for selected cases in which
>         the inner and outer header values don't match.

Well, it's not that they don't match, it's that the combination isn't 
expected to arise - it's currently unused. But I know what you meant.

>- An Appendix that discusses PCN as an example points out that
>         for PCN, it is appropriate to raise alarms for c) and d).

I think what you mean is: if PCN weren't being used, it would be _as_ 
appropriate to raise an alarm for b) as it is for c) & d).


>As for "How does this have anything to do with PCN?", the answer
>to that question is the same as the answer to "Why does the
>draft not propose an optional alarm for case b)?" ...  the
>answer is probably some form of "no current use", in which case,
>what's the current use for the c) and d) alarms? ... it's PCN.

Because I'm trying to find the lowest common denominator, I say raise 
alarms for c) & d) whether using PCN or not.

Then the question is, "Should we alarm on b) if PCN is _not_ being used?"

Is this what we really disagree on? If so, we could say:
"MAY alarm on c) & d). MAY also alarm on b) if certain that no 
3-level scheme is being used through the tunnel, but NOT RECOMMENDED 
due to the extra complexity of finding out whether a 3-level scheme is in use."

Does this help? Or have I still mis-estimated where you're coming from?


Bob


>Thanks,
>--David
>
>
> > -----Original Message-----
> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > Sent: Friday, July 10, 2009 1:32 PM
> > To: Black, David
> > Cc: tsvwg@ietf.org
> > Subject: RE: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> >
> > David,
> >
> > I'm not so pessimistic about reaching consensus on the list. This is
> > all about the 3 "nice to have" things (a,c,d). As I've already said,
> > none are anything to do with PCN. All the controversial combinations
> > are currently unused by ECN and currently unused by PCN.
> >
> > Our disagreement is over this abstract principle:
> > When changing something that's needed, is it worse or better to leave
> > other things unchanged that would be "nice to have" (ie. things that
> > are closing off  future evolvability)?
> >
> > You say it's better to leave them unchanged. Why?
> >
> > I say it's worse to leave them unchanged, because you get a big
> > potential future gain for minimal extra effort. It costs diddly-squat
> > to make 4 changes to a code block if you're already going to make 1
> > change. It requires nearly the same testing, etc etc. In return you
> > get a greater chance of 3 extra usable codepoint combinations
> > in the IP header.
> >
> > I look at the 5 different ways that ECN tunnelling has gone through
> > in the past, and I think "if we're going to add a 6th, let's
> > avoid a 7th".
> >
> >
> >
> > Bob
> >
> > At 17:09 10/07/2009, Black_David@emc.com wrote:
> > >Quick summary of the four changes proposed in the draft:
> > >
> > >a) Inner: Not-ECT, Outer: ECT(0)
> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> > >b) Inner: ECT(0), Outer: ECT(1)
> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
> > >c) Inner: ECT(1), Outer: ECT(0)
> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > >d) Inner: CE, Outer: ECT(1)
> > >         RFC 3168/4301: CE, Change: CE(!!!)
> > >
> > >If we were designing from scratch, all of these changes would
> > >be fine things to do.  We're not, and especially after the
> > >discussion of how this impacts RFC 2003 gets added, this
> > >draft will be proposing changes to some of the most widely
> > >deployed mechanisms in the Internet.  To me, that strongly
> > >suggests a conservative design approach - focus on what really
> > >needs to be changed, and resist them temptation to fix other
> > >things that would be "nice to have".  In other words, IMHO
> > >the appropriate criteria for "fixing it" is a very strong
> > >cast that "it's broken".
> > >
> > >This is an area of principled disagreement on goals between
> > >Bob and myself that list discussion is unlikely to resolve.
> > >My view of the above four changes is that only b) makes it
> > >over the "really needs to be changed" bar (and even then,
> > >only with a *much* stronger statement of the PCN-based
> > >rationale than is in the current draft, and Bob concurs
> > >with strengthening the rationale).
> > >
> > >Turning to the other three changes, a) is clearly described
> > >as "nice to have", enough said.
> > >
> > >For c) and d), I don't believe the optional alarms are
> > >justified based on ECN (RFC 3168) alone.  A key observation
> > >is that this draft does not make the < b) ECT(0) [Inner]
> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> > >optional alarm.  What this means is that this draft not only
> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
> > >*assumes* that they are ordered in terms of priority,
> > >severity, or something analogous.  That is unwarranted as
> > >a general design assumption, IMHO.
> > >
> > >For this, reason, I think discussion of which transitions are
> > >optionally alarm-able in this fashion belongs in the context of
> > >a protocol that actually orders these codepoints, not in a
> > >general document that would serve to "just constrain future
> > >standards," borrowing Toby's words ;-).
> > >
> > >I could see discussion of optional alarms in an Appendix that
> > >discusses 3-valued PCN as an example, but not in the main body
> > >of the draft.  In the main body, it would be fine to say that
> > >specific uses of the ECN codepoints may assign meanings that
> > >cause some of these transitions to be appropriate for alarms
> > >(and point to the Appendix for a PCN example), but it would
> > >not be ok to constrain the assignment of those meanings by
> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> > >optionally alarm-able.
> > >
> > >Thanks,
> > >--David
> > >
> > >
> > > > -----Original Message-----
> > > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > > > On Behalf Of Bob Briscoe
> > > > Sent: Thursday, July 09, 2009 3:08 PM
> > > > To: Black, David
> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > > > Subject: Re: [tsvwg] review of
> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > > >
> > > > David,
> > > >
> > > > Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...
> > > >
> > > > Your comments get to the nub of the proposed egress
> > protocol changes.
> > > > Great... inline...
> > > >
> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > > > >Bob,
> > > > >
> > > > >Sorry for the delay on this: things got entirely too
> > busy around the
> > > > >end of the quarter here.  Having re-read the draft from
> > scratch, I
> > > > >have a number of things to comment on ... I will be in Stockholm,
> > > > >and some of these items are going to need further discussion
> > > > >(private and/or public) in Stockholm.
> > > > >
> > > > >Starting with this table in Figure 4:
> > > > >
> > > >
> > >+------------------+----------------------------------------------+
> > > > >|  Incoming Inner  |             Incoming Outer Header
> >          |
> > > > >|      Header
> > +---------+------------+------------+----------+
> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
> >    CE    |
> > > >
> > >+------------------+---------+------------+------------+----------+
> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
> > drop(!!!)|
> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
> >   CE     |
> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
> >   CE     |
> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
> >   CE     |
> > > >
> > >+------------------+---------+------------+------------+----------+
> > > > >                    |                Outgoing Header
> >           |
> > > > >
> > +----------------------------------------------+
> > > > >
> > > > >My initial focus is on what needs to change from RFC
> > 3168/RFC 4301
> > > > >behavior and why.  Given that these two RFCs line up, I
> > think changes
> > > > >need strong motivation.  So, here are the four changes, and
> > > > my thoughts:
> > > > >
> > > > >a) Inner: Not-ECT, Outer: ECT(0)
> > > > >         3168/4301: drop (!!!)
> > > > >         Change: Not-ECT (!!!)
> > > > >         Rationale: "allow future standards actions to use this
> > > > >                 combination"
> > > > >         My view: This ain't broke, why are you fixing it?
> > > >
> > > > It is broke in the sense tunnel egresses unnecessarily
> > drop harmless
> > > > packets merely because they carry a "currently unused" (CU)
> > > > combination of codepoints. This is similar to the problem
> > ECN faced
> > > > with firewalls dropping packets unless they were Not-ECT.
> > > >
> > > > (a) wasn't originally part of the changes I was
> > proposing. But Anil
> > > > Agawaal convinced me it is worth doing. Reasoning: If we
> > are making a
> > > > change to decap anyway, let's get it as right as possible. If we
> > > > decide not to make any other changes to decap, I wouldn't
> > fight for
> > > > this one alone.
> > > >
> > > >
> > > > >b) Inner: ECT(0), Outer: ECT(1)
> > > > >         3168/4301: ECT(0)
> > > > >         Change: ECT(1)
> > > > >         Rationale: Treat ECT(1) as more severe congestion
> > > > than ECT(0).
> > > > >         My view: Longer discussion below.
> > > > >
> > > > >c) Inner: ECT(1), Outer: ECT(0)
> > > > >         3168/4301: ECT(1)
> > > > >         Change: ECT(1) (!!!)
> > > > >         Rationale: "no currently legal transition" can
> > > > produce this combo.
> > > > >         My view: Not a good idea, among other reasons
> > > > because it's at
> > > > >                 odds with your "future standards action"
> > > > explanation for
> > > > >                 the first change.  I'd remove this change
> > > > and instead deal
> > > > >                 with the possibility of raising an alarm in
> > > > a PCN-specific
> > > > >                 draft rather than declaring this to be an
> > > > alarm-able combo
> > > > >                 in general.
> > > > >
> > > > >d) Inner: CE, Outer: ECT(1)
> > > > >         3168/4301: CE
> > > > >         Change: CE (!!!)
> > > > >         Rationale & My view: <same as previous change>
> > > >
> > > > c) & d) aren't PCN specific.
> > > >
> > > > Reasoning for adding alarms (also said this eariler on list in
> > > > conversation with Gorry):
> > > > If security folks have concerns over possible attacks
> > from "currently
> > > > unused" combinations, I want to give them a way to manage the
> > > > perceived threat without just dropping packets. RFC3168
> > recommended
> > > > alarms for CU combinations. In these cases, I believe
> > RFC3168 simply
> > > > overlooked that these are also CU combinations.
> > > >
> > > >
> > > > >So, I would remove a) as not defensible (no good reason
> > for change),
> > > > >and I would push the discussion of adding alarms for c)
> > and d) into
> > > > >a PCN-specific document.
> > > >
> > > > c) & d) are not to do with PCN.
> > > >
> > > > My motivation is: While making change (b) to enable PCN, lets do
> > > > (a,c,d) to make the IP layer as clean as possible, to
> > ensure easier
> > > > future evolvability.
> > > >
> > > > I don't see the issue. It says SHOULD log and MAY alarm. So the
> > > > operator doesn't have to do anything.
> > > >
> > > > >The c) alarm in particular is only
> > > > >meaningful if PCN or something like it is in use.
> > > >
> > > > No. The combination in c) always was CU, and it's still
> > CU with PCN.
> > > > None of the proposed changes are PCN-specific.
> > > >
> > > > Even (b) is intended to also be useful for any 3-level
> > ECN while not
> > > > precluding 2-level ECN and the nonce.
> > > >
> > > > NB:  3 level = 2 'congested' levels + 1 level not
> > congested, e.g. VCP
> > > > or PCN which were both mentioned in RFC4774. This doesn't
> > count the
> > > > not-ECN-capable state, which makes 4 states.
> > > >
> > > > >I don't view d)
> > > > >as deserving of an alarm in general, as even though something has
> > > > >happened, no damage has been done - this is in line with the
> > > > >"be liberal in what you accept" principle of protocol design.
> > > >
> > > > That's why I'm proposing an optional alarm. Optional alarms *are*
> > > > liberal, but they allow operators a way to address
> > security concerns
> > > > about 'known unknowns' [Rumsfelt], while still allowing new
> > > > behaviours to be introduced.
> > > >
> > > >
> > > > >That leaves b), which I think is the key desired change, because
> > > > >it makes a third ECN code point usable through tunnels.
> > > >
> > > > Yes - exactly.
> > > >
> > > > >Much as
> > > > >this might be useful for PCN, it may be too late to do this.  The
> > > > >problem starts from the need to say that this draft
> > updates RFC 4301.
> > > > >The result will be a mix of 4301 and updated-4301 implementations
> > > > >that render the third state difficult to use in practice.  The
> > > > >problem's even worse for non-IPsec tunnels.
> > > >
> > > > The idea is that an operator wanting to implement PCN will have a
> > > > spec they can point to in their invitation to tender. At the same
> > > > time this spec both UPDATEs and is backward compatible
> > with RFC3168 &
> > > > RFC4301 because it only changes CU stuff at egress and
> > the change at
> > > > the ingress leads to the same output at the egress.
> > Therefore vendors
> > > > can update 3168 & 4301 tunnels lazily - only when their
> > customers ask
> > > > for the new features (whether for PCN or for something
> > else like VCP).
> > > >
> > > > >I also read Appendix D
> > > > >as saying that PCN is going the two-state route, not three.
> > > >
> > > > That's not how it should be read - I'll clarify it. The
> > PCN w-g wants
> > > > 3 states, but can't have them while ecn-tunnel is not proposed
> > > > standard. Therefore, it has decided to define a 2-state
> > baseline and
> > > > a number of experimental schemes to work round this tunnel issue
> > > > (plus a comparison doc about them all). It has agreed
> > that eventually
> > > > just one experimental scheme will be chosen for standards track:
> > > > - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> > > > - 3-state (which uses an extra DSCP)
> > > > - psdm (which requires out of band state)
> > > > - sm (single marking gets 3 functions from 2 encoding states but
> > > > admits admission control & termination aren't very precise in
> > > > all cases)
> > > >
> > > > ___
> > > > See part 2/2 for responses to your comments on specific parts
> > > > of the test.
> > > >
> > > > Thanks again for the review
> > > >
> > > >
> > > > Bob
> > > >
> > > >
> > > >
> > > > ________________________________________________________________
> > > > Bob Briscoe,               Networks Research Centre, BT Research
> > > >
> > > >
> > > >
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From Black_David@emc.com  Mon Jul 13 05:54:17 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 550AF28C2E3 for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 05:54:17 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.402
X-Spam-Level: 
X-Spam-Status: No, score=-6.402 tagged_above=-999 required=5 tests=[AWL=0.197,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id eyF7Y86Fo7kg for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 05:54:16 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id 6D0BA3A6DAA for <tsvwg@ietf.org>; Mon, 13 Jul 2009 05:53:50 -0700 (PDT)
Received: from hop04-l1d11-si03.isus.emc.com (HOP04-L1D11-SI03.isus.emc.com [10.254.111.23]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6DCsG0h007975 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Mon, 13 Jul 2009 08:54:16 -0400
Received: from mailhub.lss.emc.com (numailhub.lss.emc.com [10.254.144.16]) by hop04-l1d11-si03.isus.emc.com (Tablus Interceptor); Mon, 13 Jul 2009 08:54:12 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2/Switch-3.3.2) with ESMTP id n6DCrwwK006452; Mon, 13 Jul 2009 08:53:58 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 13 Jul 2009 08:53:58 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Mon, 13 Jul 2009 08:53:57 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
Thread-Index: AcoDqh7kxgH4qPkZSl2xWoqmrT4arwADV5iw
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk><9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com><200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk><9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 13 Jul 2009 12:53:58.0603 (UTC) FILETIME=[FCF179B0:01CA03B8]
X-EMM-EM: Active
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 13 Jul 2009 12:54:17 -0000

Bob,

> We're getting somewhere (if we set aside (a) for now).
>
> I'll first try to reverse engineer how the way we think differs...
> - You seem to be thinking of different ECN tunnel behaviours for=20
> different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> - My goal with this draft is to find a lowest common denominator ECN=20
> tunnel behaviour for all types of tunnel (IPsec, non-IPsec, 3-level,=20
> 2-level, ECN not supported, whatever).

We're actually aligned on this - I agree with you that the latter is
the goal.  However, we disagree on what the lowest common denominator
is ...

[... snip ...]

> Because I'm trying to find the lowest common denominator, I say raise=20
> alarms for c) & d) whether using PCN or not.

If PCN is not in use, alarms on c) and d) are not needed, IMHO
(e.g., cf. RFC 3168).

I believe no alarms on c) and d) is the lowest common denominator.
What this means is that the baseline (lowest common denominator)
alarms only when a not-ECT Inner packet comes through with an ECN
codepoint in Outer header.

I'd state that the baseline is no alarms added, and then PCN-specific
text can then explain why for PCN:
- c) and d) are alarming, but
- b) is expected behavior.

> Then the question is, "Should we alarm on b) if PCN is _not_=20
> being used?"
>=20
> Is this what we really disagree on?

No - IMHO, we should not alarm on b), c) or d) unless something
like PCN is being used.

Thanks,
--David

> -----Original Message-----
> From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]=20
> On Behalf Of Bob Briscoe
> Sent: Monday, July 13, 2009 7:07 AM
> To: Black, David
> Cc: tsvwg@ietf.org
> Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>=20
> David,
>=20
> We're getting somewhere (if we set aside (a) for now).
>=20
> I'll first try to reverse engineer how the way we think differs...
> - You seem to be thinking of different ECN tunnel behaviours for=20
> different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> - My goal with this draft is to find a lowest common denominator ECN=20
> tunnel behaviour for all types of tunnel (IPsec, non-IPsec, 3-level,=20
> 2-level, ECN not supported, whatever).
>=20
> This separation is important in both directions:
> - all nodes should be able to expect the ECN field to behave in the=20
> same way whether or not tunnels intervene without their control
> - all tunnels should be able to process the ECN field in the same=20
> way, whatever ECN is being used for, because tunnels get deployed=20
> independently of congestion control schemes like ECN & PCN.
>=20
> However, at encap I've had to accept that just one mode is=20
> impossible, as we still need the two RFC3168 behaviours for safety=20
> with legacy egresses.
>=20
> now inline...
>=20
> At 23:28 10/07/2009, Black_David@emc.com wrote:
> >Bob,
> >
> >Actually it's only about a) if you can accept the following
> >approach for the optional alarms in c) and d):
> >
> >- Main body of draft says that specific uses of ECN may raise
> >         optional decapsulation alarms for selected cases in which
> >         the inner and outer header values don't match.
>=20
> Well, it's not that they don't match, it's that the combination isn't=20
> expected to arise - it's currently unused. But I know what you meant.
>=20
> >- An Appendix that discusses PCN as an example points out that
> >         for PCN, it is appropriate to raise alarms for c) and d).
>=20
> I think what you mean is: if PCN weren't being used, it would be _as_=20
> appropriate to raise an alarm for b) as it is for c) & d).
>=20
>=20
> >As for "How does this have anything to do with PCN?", the answer
> >to that question is the same as the answer to "Why does the
> >draft not propose an optional alarm for case b)?" ...  the
> >answer is probably some form of "no current use", in which case,
> >what's the current use for the c) and d) alarms? ... it's PCN.
>=20
> Because I'm trying to find the lowest common denominator, I say raise=20
> alarms for c) & d) whether using PCN or not.
>=20
> Then the question is, "Should we alarm on b) if PCN is _not_=20
> being used?"
>=20
> Is this what we really disagree on? If so, we could say:
> "MAY alarm on c) & d). MAY also alarm on b) if certain that no=20
> 3-level scheme is being used through the tunnel, but NOT RECOMMENDED=20
> due to the extra complexity of finding out whether a 3-level=20
> scheme is in use."
>=20
> Does this help? Or have I still mis-estimated where you're=20
> coming from?
>=20
>=20
> Bob
>=20

[... snip ...]

From rbriscoe@jungle.bt.co.uk  Mon Jul 13 10:31:30 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id D36FE3A6EBD for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 10:31:30 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.157
X-Spam-Level: 
X-Spam-Status: No, score=-2.157 tagged_above=-999 required=5 tests=[AWL=-0.040, BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id YaJnx5R-8dDn for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 10:31:24 -0700 (PDT)
Received: from smtp1.smtp.bt.com (smtp1.smtp.bt.com [217.32.164.137]) by core3.amsl.com (Postfix) with ESMTP id B800228C573 for <tsvwg@ietf.org>; Mon, 13 Jul 2009 10:30:13 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp1.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 13 Jul 2009 18:30:43 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 13 Jul 2009 18:30:43 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247506242854; Mon, 13 Jul 2009 18:30:42 +0100
Received: from mut.jungle.bt.co.uk ([10.73.131.93]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6DHUWHa000642; Mon, 13 Jul 2009 18:30:33 +0100
Message-Id: <200907131730.n6DHUWHa000642@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 13 Jul 2009 18:30:08 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 13 Jul 2009 17:30:43.0343 (UTC) FILETIME=[A623B9F0:01CA03DF]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 13 Jul 2009 17:31:31 -0000

David,

Optional alarms c) & d) have no more effect on PCN than on anything 
else, and they have no more impact on anything else whether or not 
PCN is there. They have absolutely nothing to do with PCN.

>c) Inner: ECT(1), Outer: ECT(0)
>         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)

In RFC3168, if you get this combination, something along the tunnel 
has changed ECT(1) to ECT(0). That implies something weird has 
happened - not necessarily dangerous but, as no known or envisaged 
protocols do this, it could turn into something dangerous if the 
tunnel operator doesn't find out what's going on  - which needs an 
(optional) alarm to trigger some action.

>d) Inner: CE, Outer: ECT(1)
>         RFC 3168/4301: CE, Change: CE(!!!)

Again, RFC3168 full functionality mode says if the inner is CE, set 
the outer to ECT(0). So ECT(1) in the outer implies a similar 
weirdness has happened as above. So ditto: MAY alarm.

Just because RFC3168 didn't say "MAY alarm" in these cases, doesn't 
mean it shouldn't have. I'm suggesting these were missed, that's all.



Bob

At 13:53 13/07/2009, Black_David@emc.com wrote:
>Bob,
>
> > We're getting somewhere (if we set aside (a) for now).
> >
> > I'll first try to reverse engineer how the way we think differs...
> > - You seem to be thinking of different ECN tunnel behaviours for
> > different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> > - My goal with this draft is to find a lowest common denominator ECN
> > tunnel behaviour for all types of tunnel (IPsec, non-IPsec, 3-level,
> > 2-level, ECN not supported, whatever).
>
>We're actually aligned on this - I agree with you that the latter is
>the goal.  However, we disagree on what the lowest common denominator
>is ...
>
>[... snip ...]
>
> > Because I'm trying to find the lowest common denominator, I say raise
> > alarms for c) & d) whether using PCN or not.
>
>If PCN is not in use, alarms on c) and d) are not needed, IMHO
>(e.g., cf. RFC 3168).
>
>I believe no alarms on c) and d) is the lowest common denominator.
>What this means is that the baseline (lowest common denominator)
>alarms only when a not-ECT Inner packet comes through with an ECN
>codepoint in Outer header.
>
>I'd state that the baseline is no alarms added, and then PCN-specific
>text can then explain why for PCN:
>- c) and d) are alarming, but
>- b) is expected behavior.
>
> > Then the question is, "Should we alarm on b) if PCN is _not_
> > being used?"
> >
> > Is this what we really disagree on?
>
>No - IMHO, we should not alarm on b), c) or d) unless something
>like PCN is being used.
>
>Thanks,
>--David
>
> > -----Original Message-----
> > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > On Behalf Of Bob Briscoe
> > Sent: Monday, July 13, 2009 7:07 AM
> > To: Black, David
> > Cc: tsvwg@ietf.org
> > Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> >
> > David,
> >
> > We're getting somewhere (if we set aside (a) for now).
> >
> > I'll first try to reverse engineer how the way we think differs...
> > - You seem to be thinking of different ECN tunnel behaviours for
> > different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> > - My goal with this draft is to find a lowest common denominator ECN
> > tunnel behaviour for all types of tunnel (IPsec, non-IPsec, 3-level,
> > 2-level, ECN not supported, whatever).
> >
> > This separation is important in both directions:
> > - all nodes should be able to expect the ECN field to behave in the
> > same way whether or not tunnels intervene without their control
> > - all tunnels should be able to process the ECN field in the same
> > way, whatever ECN is being used for, because tunnels get deployed
> > independently of congestion control schemes like ECN & PCN.
> >
> > However, at encap I've had to accept that just one mode is
> > impossible, as we still need the two RFC3168 behaviours for safety
> > with legacy egresses.
> >
> > now inline...
> >
> > At 23:28 10/07/2009, Black_David@emc.com wrote:
> > >Bob,
> > >
> > >Actually it's only about a) if you can accept the following
> > >approach for the optional alarms in c) and d):
> > >
> > >- Main body of draft says that specific uses of ECN may raise
> > >         optional decapsulation alarms for selected cases in which
> > >         the inner and outer header values don't match.
> >
> > Well, it's not that they don't match, it's that the combination isn't
> > expected to arise - it's currently unused. But I know what you meant.
> >
> > >- An Appendix that discusses PCN as an example points out that
> > >         for PCN, it is appropriate to raise alarms for c) and d).
> >
> > I think what you mean is: if PCN weren't being used, it would be _as_
> > appropriate to raise an alarm for b) as it is for c) & d).
> >
> >
> > >As for "How does this have anything to do with PCN?", the answer
> > >to that question is the same as the answer to "Why does the
> > >draft not propose an optional alarm for case b)?" ...  the
> > >answer is probably some form of "no current use", in which case,
> > >what's the current use for the c) and d) alarms? ... it's PCN.
> >
> > Because I'm trying to find the lowest common denominator, I say raise
> > alarms for c) & d) whether using PCN or not.
> >
> > Then the question is, "Should we alarm on b) if PCN is _not_
> > being used?"
> >
> > Is this what we really disagree on? If so, we could say:
> > "MAY alarm on c) & d). MAY also alarm on b) if certain that no
> > 3-level scheme is being used through the tunnel, but NOT RECOMMENDED
> > due to the extra complexity of finding out whether a 3-level
> > scheme is in use."
> >
> > Does this help? Or have I still mis-estimated where you're
> > coming from?
> >
> >
> > Bob
> >
>
>[... snip ...]

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From rbriscoe@jungle.bt.co.uk  Mon Jul 13 11:48:37 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 2EF933A6800 for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 11:48:37 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.154
X-Spam-Level: 
X-Spam-Status: No, score=-2.154 tagged_above=-999 required=5 tests=[AWL=-0.037, BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id QEuHz5X+P-T7 for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 11:48:31 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id 2279B3A67AD for <tsvwg@ietf.org>; Mon, 13 Jul 2009 11:48:30 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 13 Jul 2009 19:49:01 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 13 Jul 2009 19:49:01 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247510940423; Mon, 13 Jul 2009 19:49:00 +0100
Received: from mut.jungle.bt.co.uk ([10.73.131.93]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6DImXMK003748; Mon, 13 Jul 2009 19:48:33 +0100
Message-Id: <200907131848.n6DImXMK003748@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 13 Jul 2009 18:20:24 +0100
To: gorry@erg.abdn.ac.uk
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <4A52F472.3040500@erg.abdn.ac.uk>
References: <4A43D532.6040906@erg.abdn.ac.uk> <200906261328.n5QDSuiM007908@bagheera.jungle.bt.co.uk> <4A52F472.3040500@erg.abdn.ac.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 13 Jul 2009 18:49:01.0255 (UTC) FILETIME=[96503570:01CA03EA]
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] Some comments on draft-ietf-tsvwg-ecn-tunnel-02]]
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 13 Jul 2009 18:48:37 -0000

Gorry,

This is for the record, to respond to your comments about the appendices ...

At 08:08 07/07/2009, Gorry Fairhurst wrote:
>Bob Briscoe wrote:
>>At 20:51 25/06/2009, Gorry Fairhurst wrote:
>>>Bob,
>>>
>>>I've just read draft-ietf-tsvwg-ecn-tunnel-02, and have some comments.
>>>Thanks for preparing this new version, which is certainly easier to read
>>>(and seems a little shorter) than the previous one, I think as I 
>>>said at the IETF, the general approach in this draft seems good to me.
>>>
>>>I've separated my comments into two parts. The first parts are comments
>>>which I'd hope you may be able to respond to. The second parts I 
>>>think may be editorial comments (but do ask questions, if you do 
>>>not agree/understand).
>>>
>>>Best wishes,
>>>
>>>Gorry

[snip more important stuff already responded to]

>>>---
>>>10) Size of Appendices.
>>>
>>>My review ran out of steam as I read the pages of appendices and I am
>>>not currently persauded that most of the text in the appendices is
>>>needed to publish the updated specification.
>>I'm also running out of steam here, as I'm meant to be going on 
>>holiday for a week now. As my wife is politely suggesting a cup of 
>>water might end up being spilled over my keyboard, I'll also be brief...
>>
>>>Section A.2:
>>>
>>>Section B seems to be related to the rationale for PCN, rather than the
>>>use of ECN over tunnels. Is this really required in this specific document?
>>>
>>>B.1 seems to be a background to ECN
>>>B.2 again reads like a paper, but the language is rather imprecise

I've completely removed Appx B.

>>>Appendix C seems OK, but is it needed?

It placated the main concern Sally Floyd had in an earlier draft. I 
plan to keep this one in there.

>>>Appendix D seems to justify the work, but since this is now adopted, is
>>>the text neccessary?

As the next stage is to take this to the Sec Area, I would like to 
leave this justification text in at this stage (they will need strong 
justification to update RFC4301). The IESG will probably have similar 
concerns. I personally prefer to keep some record of justification 
for a change, even in a stds track RFC.

>>>Section D.1 could be removed to improve the document.

Ditto.

Cheers


Bob



________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From rbriscoe@jungle.bt.co.uk  Mon Jul 13 13:14:53 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 67E7C28C49F for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 13:14:53 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.152
X-Spam-Level: 
X-Spam-Status: No, score=-2.152 tagged_above=-999 required=5 tests=[AWL=-0.035, BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 8Sgb66dUOkOa for <tsvwg@core3.amsl.com>; Mon, 13 Jul 2009 13:14:47 -0700 (PDT)
Received: from smtp4.smtp.bt.com (smtp4.smtp.bt.com [217.32.164.151]) by core3.amsl.com (Postfix) with ESMTP id 35B943A69FC for <tsvwg@ietf.org>; Mon, 13 Jul 2009 13:14:41 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp4.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 13 Jul 2009 21:15:11 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 13 Jul 2009 21:15:10 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1247516109616; Mon, 13 Jul 2009 21:15:09 +0100
Received: from mut.jungle.bt.co.uk ([10.73.131.93]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6DKF5SJ007774; Mon, 13 Jul 2009 21:15:06 +0100
Message-Id: <200907132015.n6DKF5SJ007774@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 13 Jul 2009 21:15:16 +0100
To: gorry@erg.abdn.ac.uk
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <4A52F472.3040500@erg.abdn.ac.uk>
References: <4A43D532.6040906@erg.abdn.ac.uk> <200906261328.n5QDSuiM007908@bagheera.jungle.bt.co.uk> <4A52F472.3040500@erg.abdn.ac.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 13 Jul 2009 20:15:10.0776 (UTC) FILETIME=[9F968380:01CA03F6]
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] Some comments on draft-ietf-tsvwg-ecn-tunnel-02]]
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 13 Jul 2009 20:14:53 -0000

Gorry,

Again for the record, here's the (two) nits I didn't do. With reasons...

At 08:08 07/07/2009, Gorry Fairhurst wrote:
>>Tx for the nits below. I'll handle them.
>>
>>Bob
>>
>>>----------------------------------------------------------------------
>>>* NiTs

>>>---
>>>1.1 Scope, para 2:
>>>- should this refer to RFC 2983 for example?

2983 is already referred to a couple of paras later, and it would be 
confusing to refer to it in this para, which states precisely the 
scope of this spec (given the scope of 2983 is very different). I've 
added a ref to 4774 at this point tho.

>>>---
>>>Section 4.3:
>>>/S.5/
>>>  ^ - Section?

I put in the section symbol (two S's on top of each other) whenever I 
xref a specific section of another RFC, and the tools output it 
correctly in the .html, and convert it to S. in the .txt. I think 
this is a sufficiently well-understood abbreviation that is also 
guessable, so I'd rather leave it.

   


From Black_David@emc.com  Tue Jul 14 20:21:23 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 863383A691B for <tsvwg@core3.amsl.com>; Tue, 14 Jul 2009 20:21:23 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.416
X-Spam-Level: 
X-Spam-Status: No, score=-6.416 tagged_above=-999 required=5 tests=[AWL=0.183,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 0QAzOVGs8Q5z for <tsvwg@core3.amsl.com>; Tue, 14 Jul 2009 20:21:22 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id F031D3A68B9 for <tsvwg@ietf.org>; Tue, 14 Jul 2009 20:21:21 -0700 (PDT)
Received: from hop04-l1d11-si02.isus.emc.com (HOP04-L1D11-SI02.isus.emc.com [10.254.111.55]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6F3JIi1029251 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Tue, 14 Jul 2009 23:19:19 -0400
Received: from mailhub.lss.emc.com (nagas.lss.emc.com [10.254.144.15]) by hop04-l1d11-si02.isus.emc.com (Tablus Interceptor); Tue, 14 Jul 2009 23:19:07 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2mp/Switch-3.3.2mp) with ESMTP id n6F3J5IP015569; Tue, 14 Jul 2009 23:19:06 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Tue, 14 Jul 2009 23:19:06 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Tue, 14 Jul 2009 23:18:59 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A033673E9@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907131730.n6DHUWHa000642@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
Thread-Index: AcoD38nFD/iMU624QtWLDx57pwoRIQBGOOzg
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp.emc.com> <200907131730.n6DHUWHa000642@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 15 Jul 2009 03:19:06.0166 (UTC) FILETIME=[02ACC560:01CA04FB]
X-EMM-EM: Active
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 15 Jul 2009 03:21:23 -0000

Bob,

> Optional alarms c) & d) have no more effect on PCN than on anything=20
> else, and they have no more impact on anything else whether or not=20
> PCN is there. They have absolutely nothing to do with PCN.

I think you've missed the point; in isolation, the above statement
is correct, ... BUT ... the triple of <no b) alarm, c) alarm,
d) alarm> is PCN-specific, AFAICT ...

> >c) Inner: ECT(1), Outer: ECT(0)
> >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
>
> In RFC3168, if you get this combination, something along the tunnel=20
> has changed ECT(1) to ECT(0). That implies something weird has=20
> happened - not necessarily dangerous but, as no known or envisaged=20
> protocols do this, it could turn into something dangerous if the=20
> tunnel operator doesn't find out what's going on  - which needs an=20
> (optional) alarm to trigger some action.

The problem with this reasoning is that wrt RFC3168, b) is the same
class of weirdness - ECT(0) has changed to ECT(1), yet no (optional)
alarm is proposed.  Keep in mind that ECT(0) and ECT (1) are *not*
ordered values in RFC3168 - they're unordered code points.  The
only escape I see is from this is that PCN is the "known or
envisaged" protocol that uses b) - is there something I've missed?

The d) discussion has the same issue, hence it's snipped.

[... snip ...]

> Just because RFC3168 didn't say "MAY alarm" in these cases, doesn't=20
> mean it shouldn't have. I'm suggesting these were missed, that's all.

And had RFC 3168 done that, it would also have suggested an optional
alarm for b).  If an optional alarm for b) is not wanted, the result
is PCN-specific and belongs in an Appendix, IMHO.  The result
enhances flexibility for future usage:
- Main body says that some unexpected changes within the tunnel MAY
	generate optional alarms, but expected ECN transitions (per
	RFC3168) SHOULD NOT generate alarms.
- Appendix discussing PCN as an example says that *if* ECT(0) and
	ECT(1) are ordered (true for PCN, false for RFC3168), *then*
	b) is expected behavior, but c) and d) deserve optional alarms.

Thanks,
--David
=20

> -----Original Message-----
> From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]=20
> Sent: Monday, July 13, 2009 1:30 PM
> To: Black, David
> Cc: tsvwg@ietf.org
> Subject: RE: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>=20
> David,
>=20
> Optional alarms c) & d) have no more effect on PCN than on anything=20
> else, and they have no more impact on anything else whether or not=20
> PCN is there. They have absolutely nothing to do with PCN.
>=20
> >c) Inner: ECT(1), Outer: ECT(0)
> >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
>=20
> In RFC3168, if you get this combination, something along the tunnel=20
> has changed ECT(1) to ECT(0). That implies something weird has=20
> happened - not necessarily dangerous but, as no known or envisaged=20
> protocols do this, it could turn into something dangerous if the=20
> tunnel operator doesn't find out what's going on  - which needs an=20
> (optional) alarm to trigger some action.
>=20
> >d) Inner: CE, Outer: ECT(1)
> >         RFC 3168/4301: CE, Change: CE(!!!)
>=20
> Again, RFC3168 full functionality mode says if the inner is CE, set=20
> the outer to ECT(0). So ECT(1) in the outer implies a similar=20
> weirdness has happened as above. So ditto: MAY alarm.
>=20
> Just because RFC3168 didn't say "MAY alarm" in these cases, doesn't=20
> mean it shouldn't have. I'm suggesting these were missed, that's all.
>=20
>=20
>=20
> Bob
>=20
> At 13:53 13/07/2009, Black_David@emc.com wrote:
> >Bob,
> >
> > > We're getting somewhere (if we set aside (a) for now).
> > >
> > > I'll first try to reverse engineer how the way we think differs...
> > > - You seem to be thinking of different ECN tunnel behaviours for
> > > different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> > > - My goal with this draft is to find a lowest common=20
> denominator ECN
> > > tunnel behaviour for all types of tunnel (IPsec,=20
> non-IPsec, 3-level,
> > > 2-level, ECN not supported, whatever).
> >
> >We're actually aligned on this - I agree with you that the latter is
> >the goal.  However, we disagree on what the lowest common denominator
> >is ...
> >
> >[... snip ...]
> >
> > > Because I'm trying to find the lowest common denominator,=20
> I say raise
> > > alarms for c) & d) whether using PCN or not.
> >
> >If PCN is not in use, alarms on c) and d) are not needed, IMHO
> >(e.g., cf. RFC 3168).
> >
> >I believe no alarms on c) and d) is the lowest common denominator.
> >What this means is that the baseline (lowest common denominator)
> >alarms only when a not-ECT Inner packet comes through with an ECN
> >codepoint in Outer header.
> >
> >I'd state that the baseline is no alarms added, and then PCN-specific
> >text can then explain why for PCN:
> >- c) and d) are alarming, but
> >- b) is expected behavior.
> >
> > > Then the question is, "Should we alarm on b) if PCN is _not_
> > > being used?"
> > >
> > > Is this what we really disagree on?
> >
> >No - IMHO, we should not alarm on b), c) or d) unless something
> >like PCN is being used.
> >
> >Thanks,
> >--David
> >
> > > -----Original Message-----
> > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > > On Behalf Of Bob Briscoe
> > > Sent: Monday, July 13, 2009 7:07 AM
> > > To: Black, David
> > > Cc: tsvwg@ietf.org
> > > Subject: Re: [tsvwg] review of=20
> draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > >
> > > David,
> > >
> > > We're getting somewhere (if we set aside (a) for now).
> > >
> > > I'll first try to reverse engineer how the way we think differs...
> > > - You seem to be thinking of different ECN tunnel behaviours for
> > > different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> > > - My goal with this draft is to find a lowest common=20
> denominator ECN
> > > tunnel behaviour for all types of tunnel (IPsec,=20
> non-IPsec, 3-level,
> > > 2-level, ECN not supported, whatever).
> > >
> > > This separation is important in both directions:
> > > - all nodes should be able to expect the ECN field to=20
> behave in the
> > > same way whether or not tunnels intervene without their control
> > > - all tunnels should be able to process the ECN field in the same
> > > way, whatever ECN is being used for, because tunnels get deployed
> > > independently of congestion control schemes like ECN & PCN.
> > >
> > > However, at encap I've had to accept that just one mode is
> > > impossible, as we still need the two RFC3168 behaviours for safety
> > > with legacy egresses.
> > >
> > > now inline...
> > >
> > > At 23:28 10/07/2009, Black_David@emc.com wrote:
> > > >Bob,
> > > >
> > > >Actually it's only about a) if you can accept the following
> > > >approach for the optional alarms in c) and d):
> > > >
> > > >- Main body of draft says that specific uses of ECN may raise
> > > >         optional decapsulation alarms for selected=20
> cases in which
> > > >         the inner and outer header values don't match.
> > >
> > > Well, it's not that they don't match, it's that the=20
> combination isn't
> > > expected to arise - it's currently unused. But I know=20
> what you meant.
> > >
> > > >- An Appendix that discusses PCN as an example points out that
> > > >         for PCN, it is appropriate to raise alarms for=20
> c) and d).
> > >
> > > I think what you mean is: if PCN weren't being used, it=20
> would be _as_
> > > appropriate to raise an alarm for b) as it is for c) & d).
> > >
> > >
> > > >As for "How does this have anything to do with PCN?", the answer
> > > >to that question is the same as the answer to "Why does the
> > > >draft not propose an optional alarm for case b)?" ...  the
> > > >answer is probably some form of "no current use", in which case,
> > > >what's the current use for the c) and d) alarms? ... it's PCN.
> > >
> > > Because I'm trying to find the lowest common denominator,=20
> I say raise
> > > alarms for c) & d) whether using PCN or not.
> > >
> > > Then the question is, "Should we alarm on b) if PCN is _not_
> > > being used?"
> > >
> > > Is this what we really disagree on? If so, we could say:
> > > "MAY alarm on c) & d). MAY also alarm on b) if certain that no
> > > 3-level scheme is being used through the tunnel, but NOT=20
> RECOMMENDED
> > > due to the extra complexity of finding out whether a 3-level
> > > scheme is in use."
> > >
> > > Does this help? Or have I still mis-estimated where you're
> > > coming from?
> > >
> > >
> > > Bob
> > >
> >
> >[... snip ...]
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research=20
>=20
>=20
>=20

From jmpolk@cisco.com  Thu Jul 16 16:42:57 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 77EA728C0D8 for <tsvwg@core3.amsl.com>; Thu, 16 Jul 2009 16:42:57 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.372
X-Spam-Level: 
X-Spam-Status: No, score=-6.372 tagged_above=-999 required=5 tests=[AWL=0.227,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id akZutI2ZaFZ1 for <tsvwg@core3.amsl.com>; Thu, 16 Jul 2009 16:42:56 -0700 (PDT)
Received: from sj-iport-1.cisco.com (sj-iport-1.cisco.com [171.71.176.70]) by core3.amsl.com (Postfix) with ESMTP id 975A93A687D for <tsvwg@ietf.org>; Thu, 16 Jul 2009 16:42:56 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: Ao0IAKdXX0qrR7PE/2dsb2JhbACIQ69+iCORDQWCOoFTgUU
X-IronPort-AV: E=Sophos;i="4.42,413,1243814400"; d="scan'208";a="215229896"
Received: from sj-dkim-4.cisco.com ([171.71.179.196]) by sj-iport-1.cisco.com with ESMTP; 16 Jul 2009 23:42:06 +0000
Received: from sj-core-4.cisco.com (sj-core-4.cisco.com [171.68.223.138]) by sj-dkim-4.cisco.com (8.12.11/8.12.11) with ESMTP id n6GNg5Is026138 for <tsvwg@ietf.org>; Thu, 16 Jul 2009 16:42:05 -0700
Received: from xbh-sjc-231.amer.cisco.com (xbh-sjc-231.cisco.com [128.107.191.100]) by sj-core-4.cisco.com (8.13.8/8.14.3) with ESMTP id n6GNg6CA007378 for <tsvwg@ietf.org>; Thu, 16 Jul 2009 23:42:06 GMT
Received: from xfe-sjc-212.amer.cisco.com ([171.70.151.187]) by xbh-sjc-231.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 16 Jul 2009 16:42:05 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.4.15]) by xfe-sjc-212.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 16 Jul 2009 16:42:05 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Thu, 16 Jul 2009 18:42:04 -0500
To: tsvwg <tsvwg@ietf.org>
From: "James M. Polk" <jmpolk@cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com>
X-OriginalArrivalTime: 16 Jul 2009 23:42:05.0617 (UTC) FILETIME=[06A62A10:01CA066F]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=1849; t=1247787726; x=1248651726; c=relaxed/simple; s=sjdkim4002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20New=20rev=20of=20IntServ=20MULTI_TSPEC |Sender:=20; bh=E45ROtqiAHisurdMuj+P6ZCNE2YgvxJgXeQXEUUll4U=; b=mmAdKGoA2JUlwNSXe8PgiGzAo8A3D8AgwmX+jtjdak9lo1rZp22p1yXAT5 47pk4n1L4YEDlRu8KRu9OYfZEQF+4cvxJfZNhq5SrF2ZbP1H5U2pJg++NMZO DwXV8c79sX;
Authentication-Results: sj-dkim-4; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim4002 verified; ); 
Subject: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 16 Jul 2009 23:42:57 -0000

TSVWG

[with my individual hat on]

We have updated the ID for the Integrated Services extension for 
multiple bandwidths in a single message.

http://www.ietf.org/internet-drafts/draft-polk-tsvwg-intserv-multiple-tspec-01.txt

As the WG wanted from the preso given at IETF75 in TSVWG...

- we focused the proposal on Option#3 (leaving the TSPEC alone, and 
creating the new MULTI_TSPEC Object for all subsequent TSPECs in the 
PATH message)

- we have addressed the RSPEC issue - concluding there needs to be an 
RSPEC per TSPEC for Guaranteed Service. I explain this and show it as 
well in the ID now.

- with above conclusion means I can't just have a MULTI_TSPEC, but I 
have to limit that object for just the PATH message, and create a 
MULTI_FLOWSPEC for the RESV message. I explain this and show it as 
well in the ID now.

- we consolidated all the rules of usage into their own section, so 
there's only 1 place to have to look for this information.

- we streamlined part of the doc to flow better.

A number of folks commented specifically about there being 3 options 
proposed forward, that once we choose 1 (and the WG did), that the 
others remain in the doc so there isn't a discussion in the years to 
come about "why we didn't do *it8* another way".  We moved all the 
text for Options #1 and 2 to the appendix, but do not know if the 
existing section 2 - which talks about the pluses and minuses of each 
solution - should be all we say (i.e., thereby reducing the length of 
the doc, given the point is probably made about the other options).

We would like feedback on that issue, and others raised in the doc.

Comments are appreciated

BTW - we plan to ask the WG if this is worth taking on as a WG item, 
given all the discussion on the list and at the SF TSVWG meeting.

James


From rbriscoe@jungle.bt.co.uk  Mon Jul 20 03:03:08 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 5320E3A6CA4 for <tsvwg@core3.amsl.com>; Mon, 20 Jul 2009 03:03:08 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.117
X-Spam-Level: 
X-Spam-Status: No, score=-2.117 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id LGWQqk1LbpIR for <tsvwg@core3.amsl.com>; Mon, 20 Jul 2009 03:03:02 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id 792AF3A6C36 for <tsvwg@ietf.org>; Mon, 20 Jul 2009 03:03:01 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 20 Jul 2009 11:01:30 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 20 Jul 2009 11:01:30 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248084089863; Mon, 20 Jul 2009 11:01:29 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6KA1Pha004613; Mon, 20 Jul 2009 11:01:26 +0100
Message-Id: <200907201001.n6KA1Pha004613@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 20 Jul 2009 11:01:16 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A033673E9@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp.emc.com> <200907131730.n6DHUWHa000642@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033673E9@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 20 Jul 2009 10:01:30.0427 (UTC) FILETIME=[0DD760B0:01CA0921]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 20 Jul 2009 10:03:08 -0000

David,

At least we now understand each other's position.
We agree a set of alarms on all of b)c)d) would not be PCN-specific.
And when looked at as a whole, missing out an alarm on b) but keeping 
alarms c)d) would be PCN-specific (or 2-severity-level-specific to be precise).

But when looking at each case on its individual merits, an optional 
alarm on c) or d) makes sense to pick up unexpected behaviours, 
whether or not b) is alarmed.

I am happy to say that all three of b)c)&d) SHOULD be logged and MAY 
be alarmed, then when I say that logging & alarms will need to be 
turned off if new standards actions use a combination, I can give PCN 
as an example where b) alarms should be turned off.

As PCN is on the standards track, this mention of an exception for 
PCN ought to be in the body, not an appendix, IMO.

Acceptable?

__________
I have also decided to remove completely the alarm states '(!!!)' 
from my table recapping RFC3168 & RFC4301. Originally, I found the 
mention of logging in RFC3168, but on a re-read I shouldn't have been 
so certain of what it meant, as the text is rather vague. Also I've 
found there's actually no mention of alarms in RFC4301. So all we 
have is this sentence in RFC3168:

"9.2.2.  Changes to the ECN Field within an IPsec Tunnel.

    If the ECN Field is changed inappropriately within an IPsec tunnel,
    and this change is detected at the tunnel egress, then the receipt of
    a packet not satisfying the appropriate condition for its SA is an
    auditable event. [...]"

It doesn't actually enumerate the "inappropriate" cases, so an 
operator could construe this as already blessing alarms on b)c)d), at 
least in the IPsec case.

[Sorry for spending a couple of days before responding...]


Bob

At 04:18 15/07/2009, Black_David@emc.com wrote:
>Bob,
>
> > Optional alarms c) & d) have no more effect on PCN than on anything
> > else, and they have no more impact on anything else whether or not
> > PCN is there. They have absolutely nothing to do with PCN.
>
>I think you've missed the point; in isolation, the above statement
>is correct, ... BUT ... the triple of <no b) alarm, c) alarm,
>d) alarm> is PCN-specific, AFAICT ...
>
> > >c) Inner: ECT(1), Outer: ECT(0)
> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> >
> > In RFC3168, if you get this combination, something along the tunnel
> > has changed ECT(1) to ECT(0). That implies something weird has
> > happened - not necessarily dangerous but, as no known or envisaged
> > protocols do this, it could turn into something dangerous if the
> > tunnel operator doesn't find out what's going on  - which needs an
> > (optional) alarm to trigger some action.
>
>The problem with this reasoning is that wrt RFC3168, b) is the same
>class of weirdness - ECT(0) has changed to ECT(1), yet no (optional)
>alarm is proposed.  Keep in mind that ECT(0) and ECT (1) are *not*
>ordered values in RFC3168 - they're unordered code points.  The
>only escape I see is from this is that PCN is the "known or
>envisaged" protocol that uses b) - is there something I've missed?
>
>The d) discussion has the same issue, hence it's snipped.
>
>[... snip ...]
>
> > Just because RFC3168 didn't say "MAY alarm" in these cases, doesn't
> > mean it shouldn't have. I'm suggesting these were missed, that's all.
>
>And had RFC 3168 done that, it would also have suggested an optional
>alarm for b).  If an optional alarm for b) is not wanted, the result
>is PCN-specific and belongs in an Appendix, IMHO.  The result
>enhances flexibility for future usage:
>- Main body says that some unexpected changes within the tunnel MAY
>         generate optional alarms, but expected ECN transitions (per
>         RFC3168) SHOULD NOT generate alarms.
>- Appendix discussing PCN as an example says that *if* ECT(0) and
>         ECT(1) are ordered (true for PCN, false for RFC3168), *then*
>         b) is expected behavior, but c) and d) deserve optional alarms.
>
>Thanks,
>--David
>
>
> > -----Original Message-----
> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > Sent: Monday, July 13, 2009 1:30 PM
> > To: Black, David
> > Cc: tsvwg@ietf.org
> > Subject: RE: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> >
> > David,
> >
> > Optional alarms c) & d) have no more effect on PCN than on anything
> > else, and they have no more impact on anything else whether or not
> > PCN is there. They have absolutely nothing to do with PCN.
> >
> > >c) Inner: ECT(1), Outer: ECT(0)
> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> >
> > In RFC3168, if you get this combination, something along the tunnel
> > has changed ECT(1) to ECT(0). That implies something weird has
> > happened - not necessarily dangerous but, as no known or envisaged
> > protocols do this, it could turn into something dangerous if the
> > tunnel operator doesn't find out what's going on  - which needs an
> > (optional) alarm to trigger some action.
> >
> > >d) Inner: CE, Outer: ECT(1)
> > >         RFC 3168/4301: CE, Change: CE(!!!)
> >
> > Again, RFC3168 full functionality mode says if the inner is CE, set
> > the outer to ECT(0). So ECT(1) in the outer implies a similar
> > weirdness has happened as above. So ditto: MAY alarm.
> >
> > Just because RFC3168 didn't say "MAY alarm" in these cases, doesn't
> > mean it shouldn't have. I'm suggesting these were missed, that's all.
> >
> >
> >
> > Bob
> >
> > At 13:53 13/07/2009, Black_David@emc.com wrote:
> > >Bob,
> > >
> > > > We're getting somewhere (if we set aside (a) for now).
> > > >
> > > > I'll first try to reverse engineer how the way we think differs...
> > > > - You seem to be thinking of different ECN tunnel behaviours for
> > > > different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> > > > - My goal with this draft is to find a lowest common
> > denominator ECN
> > > > tunnel behaviour for all types of tunnel (IPsec,
> > non-IPsec, 3-level,
> > > > 2-level, ECN not supported, whatever).
> > >
> > >We're actually aligned on this - I agree with you that the latter is
> > >the goal.  However, we disagree on what the lowest common denominator
> > >is ...
> > >
> > >[... snip ...]
> > >
> > > > Because I'm trying to find the lowest common denominator,
> > I say raise
> > > > alarms for c) & d) whether using PCN or not.
> > >
> > >If PCN is not in use, alarms on c) and d) are not needed, IMHO
> > >(e.g., cf. RFC 3168).
> > >
> > >I believe no alarms on c) and d) is the lowest common denominator.
> > >What this means is that the baseline (lowest common denominator)
> > >alarms only when a not-ECT Inner packet comes through with an ECN
> > >codepoint in Outer header.
> > >
> > >I'd state that the baseline is no alarms added, and then PCN-specific
> > >text can then explain why for PCN:
> > >- c) and d) are alarming, but
> > >- b) is expected behavior.
> > >
> > > > Then the question is, "Should we alarm on b) if PCN is _not_
> > > > being used?"
> > > >
> > > > Is this what we really disagree on?
> > >
> > >No - IMHO, we should not alarm on b), c) or d) unless something
> > >like PCN is being used.
> > >
> > >Thanks,
> > >--David
> > >
> > > > -----Original Message-----
> > > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > > > On Behalf Of Bob Briscoe
> > > > Sent: Monday, July 13, 2009 7:07 AM
> > > > To: Black, David
> > > > Cc: tsvwg@ietf.org
> > > > Subject: Re: [tsvwg] review of
> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > > >
> > > > David,
> > > >
> > > > We're getting somewhere (if we set aside (a) for now).
> > > >
> > > > I'll first try to reverse engineer how the way we think differs...
> > > > - You seem to be thinking of different ECN tunnel behaviours for
> > > > different types of tunnel (IPsec, non-IPsec, PCN, non-PCN etc).
> > > > - My goal with this draft is to find a lowest common
> > denominator ECN
> > > > tunnel behaviour for all types of tunnel (IPsec,
> > non-IPsec, 3-level,
> > > > 2-level, ECN not supported, whatever).
> > > >
> > > > This separation is important in both directions:
> > > > - all nodes should be able to expect the ECN field to
> > behave in the
> > > > same way whether or not tunnels intervene without their control
> > > > - all tunnels should be able to process the ECN field in the same
> > > > way, whatever ECN is being used for, because tunnels get deployed
> > > > independently of congestion control schemes like ECN & PCN.
> > > >
> > > > However, at encap I've had to accept that just one mode is
> > > > impossible, as we still need the two RFC3168 behaviours for safety
> > > > with legacy egresses.
> > > >
> > > > now inline...
> > > >
> > > > At 23:28 10/07/2009, Black_David@emc.com wrote:
> > > > >Bob,
> > > > >
> > > > >Actually it's only about a) if you can accept the following
> > > > >approach for the optional alarms in c) and d):
> > > > >
> > > > >- Main body of draft says that specific uses of ECN may raise
> > > > >         optional decapsulation alarms for selected
> > cases in which
> > > > >         the inner and outer header values don't match.
> > > >
> > > > Well, it's not that they don't match, it's that the
> > combination isn't
> > > > expected to arise - it's currently unused. But I know
> > what you meant.
> > > >
> > > > >- An Appendix that discusses PCN as an example points out that
> > > > >         for PCN, it is appropriate to raise alarms for
> > c) and d).
> > > >
> > > > I think what you mean is: if PCN weren't being used, it
> > would be _as_
> > > > appropriate to raise an alarm for b) as it is for c) & d).
> > > >
> > > >
> > > > >As for "How does this have anything to do with PCN?", the answer
> > > > >to that question is the same as the answer to "Why does the
> > > > >draft not propose an optional alarm for case b)?" ...  the
> > > > >answer is probably some form of "no current use", in which case,
> > > > >what's the current use for the c) and d) alarms? ... it's PCN.
> > > >
> > > > Because I'm trying to find the lowest common denominator,
> > I say raise
> > > > alarms for c) & d) whether using PCN or not.
> > > >
> > > > Then the question is, "Should we alarm on b) if PCN is _not_
> > > > being used?"
> > > >
> > > > Is this what we really disagree on? If so, we could say:
> > > > "MAY alarm on c) & d). MAY also alarm on b) if certain that no
> > > > 3-level scheme is being used through the tunnel, but NOT
> > RECOMMENDED
> > > > due to the extra complexity of finding out whether a 3-level
> > > > scheme is in use."
> > > >
> > > > Does this help? Or have I still mis-estimated where you're
> > > > coming from?
> > > >
> > > >
> > > > Bob
> > > >
> > >
> > >[... snip ...]
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From rbriscoe@jungle.bt.co.uk  Mon Jul 20 04:48:32 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 35BEA3A68A0 for <tsvwg@core3.amsl.com>; Mon, 20 Jul 2009 04:48:32 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.288
X-Spam-Level: 
X-Spam-Status: No, score=-2.288 tagged_above=-999 required=5 tests=[AWL=1.829,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 6z6xPbf1U-n3 for <tsvwg@core3.amsl.com>; Mon, 20 Jul 2009 04:48:26 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id 0EEBD3A6A8F for <tsvwg@ietf.org>; Mon, 20 Jul 2009 04:48:25 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 20 Jul 2009 12:48:23 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 20 Jul 2009 12:48:23 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248090503746; Mon, 20 Jul 2009 12:48:23 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6KBmHDt009985; Mon, 20 Jul 2009 12:48:17 +0100
Message-Id: <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 20 Jul 2009 12:48:04 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 20 Jul 2009 11:48:23.0554 (UTC) FILETIME=[FC5CE620:01CA092F]
Cc: tsvwg@ietf.org
Subject: [tsvwg] draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 20 Jul 2009 11:48:32 -0000

David,

I wanted to move on from optional alarms to my proposed functional 
changes, ie. your case a).

But I have to admit a big error which makes that moot. In 
ecn-tunnel-02 I incorrectly reported what RFC3168 & RFC4301 said (I 
just re-read them). See the table below.

IP in IP Decapsulation; Recap of Pre-existing Behaviour
+------------------+----------------------------------------------+
|  Incoming Inner  |             Incoming Outer Header            |
|      Header      +---------+------------+------------+----------+
|                  | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
+------------------+---------+------------+------------+----------+
|     Not-ECT      | Not-ECT | drop       | drop       | drop     |<my error
|     Not-ECT      | Not-ECT | Not-ECT    | Not-ECT    | drop     |<-3168
|     Not-ECT      | Not-ECT | Not-ECT    | Not-ECT    | Not-ECT  |<-4301
|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
|      ECT(1)      |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
|        CE        |      CE |     CE     |     CE     |   CE     |
+------------------+---------+------------+------------+----------+
                    |                Outgoing Header               |
                    +----------------------------------------------+

RFC4301 seems to say: If the inner is Not-ECT, anything else in the 
outer must be an error. But whereas RFC3168 dropped the packet if 
there was CE in the outer, RFC4301 says forward it as Not-ECT in all cases.

RFC4301 gives no justification for this rather important change from 
RFC3168. I've spent the last hour or so trawling the IPsec and tsvwg 
archives and can't find any discussion on this point either. Was it 
discussed anywhere?

This RFC4301 behaviour seems unsafe, but perhaps understandable. It 
all depends how the situation arose:
* If the ingress wrongly turned a Not-ECT inner into an ECT outer, 
then subsequently a congested router marked it CE, then it's unsafe 
to remove CE and forward Not-ECT - drop would have been safer.
* If something broken set CE on the outer even tho there was no 
congestion, it's safe.



Bob

At 12:07 13/07/2009, Bob Briscoe wrote:
>David,
>
>We're getting somewhere (if we set aside (a) for now).
>
> > At 17:09 10/07/2009, Black_David@emc.com wrote:
>> > >Quick summary of the four changes proposed in the draft:
>> > >
>> > >a) Inner: Not-ECT, Outer: ECT(0)
>> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
>> > >b) Inner: ECT(0), Outer: ECT(1)
>> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
>> > >c) Inner: ECT(1), Outer: ECT(0)
>> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
>> > >d) Inner: CE, Outer: ECT(1)
>> > >         RFC 3168/4301: CE, Change: CE(!!!)
>> > >
>> > >If we were designing from scratch, all of these changes would
>> > >be fine things to do.  We're not, and especially after the
>> > >discussion of how this impacts RFC 2003 gets added, this
>> > >draft will be proposing changes to some of the most widely
>> > >deployed mechanisms in the Internet.  To me, that strongly
>> > >suggests a conservative design approach - focus on what really
>> > >needs to be changed, and resist them temptation to fix other
>> > >things that would be "nice to have".  In other words, IMHO
>> > >the appropriate criteria for "fixing it" is a very strong
>> > >cast that "it's broken".
>> > >
>> > >This is an area of principled disagreement on goals between
>> > >Bob and myself that list discussion is unlikely to resolve.
>> > >My view of the above four changes is that only b) makes it
>> > >over the "really needs to be changed" bar (and even then,
>> > >only with a *much* stronger statement of the PCN-based
>> > >rationale than is in the current draft, and Bob concurs
>> > >with strengthening the rationale).
>> > >
>> > >Turning to the other three changes, a) is clearly described
>> > >as "nice to have", enough said.
>> > >
>> > >For c) and d), I don't believe the optional alarms are
>> > >justified based on ECN (RFC 3168) alone.  A key observation
>> > >is that this draft does not make the < b) ECT(0) [Inner]
>> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
>> > >optional alarm.  What this means is that this draft not only
>> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
>> > >*assumes* that they are ordered in terms of priority,
>> > >severity, or something analogous.  That is unwarranted as
>> > >a general design assumption, IMHO.
>> > >
>> > >For this, reason, I think discussion of which transitions are
>> > >optionally alarm-able in this fashion belongs in the context of
>> > >a protocol that actually orders these codepoints, not in a
>> > >general document that would serve to "just constrain future
>> > >standards," borrowing Toby's words ;-).
>> > >
>> > >I could see discussion of optional alarms in an Appendix that
>> > >discusses 3-valued PCN as an example, but not in the main body
>> > >of the draft.  In the main body, it would be fine to say that
>> > >specific uses of the ECN codepoints may assign meanings that
>> > >cause some of these transitions to be appropriate for alarms
>> > >(and point to the Appendix for a PCN example), but it would
>> > >not be ok to constrain the assignment of those meanings by
>> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
>> > >optionally alarm-able.
>> > >
>> > >Thanks,
>> > >--David
>> > >
>> > >
>> > > > -----Original Message-----
>> > > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
>> > > > On Behalf Of Bob Briscoe
>> > > > Sent: Thursday, July 09, 2009 3:08 PM
>> > > > To: Black, David
>> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
>> > > > Subject: Re: [tsvwg] review of
>> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>> > > >
>> > > > David,
>> > > >
>> > > > Bringing to the list as agreed. I'll reply in 2 parts. Part 1/2...
>> > > >
>> > > > Your comments get to the nub of the proposed egress
>> > protocol changes.
>> > > > Great... inline...
>> > > >
>> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
>> > > > >Bob,
>> > > > >
>> > > > >Sorry for the delay on this: things got entirely too
>> > busy around the
>> > > > >end of the quarter here.  Having re-read the draft from
>> > scratch, I
>> > > > >have a number of things to comment on ... I will be in Stockholm,
>> > > > >and some of these items are going to need further discussion
>> > > > >(private and/or public) in Stockholm.
>> > > > >
>> > > > >Starting with this table in Figure 4:
>> > > > >
>> > > >
>> > >+------------------+----------------------------------------------+
>> > > > >|  Incoming Inner  |             Incoming Outer Header
>> >          |
>> > > > >|      Header
>> > +---------+------------+------------+----------+
>> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
>> >    CE    |
>> > > >
>> > >+------------------+---------+------------+------------+----------+
>> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
>> > drop(!!!)|
>> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
>> >   CE     |
>> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
>> >   CE     |
>> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
>> >   CE     |
>> > > >
>> > >+------------------+---------+------------+------------+----------+
>> > > > >                    |                Outgoing Header
>> >           |
>> > > > >
>> > +----------------------------------------------+
>> > > > >
>> > > > >My initial focus is on what needs to change from RFC
>> > 3168/RFC 4301
>> > > > >behavior and why.  Given that these two RFCs line up, I
>> > think changes
>> > > > >need strong motivation.  So, here are the four changes, and
>> > > > my thoughts:
>> > > > >
>> > > > >a) Inner: Not-ECT, Outer: ECT(0)
>> > > > >         3168/4301: drop (!!!)
>> > > > >         Change: Not-ECT (!!!)
>> > > > >         Rationale: "allow future standards actions to use this
>> > > > >                 combination"
>> > > > >         My view: This ain't broke, why are you fixing it?
>> > > >
>> > > > It is broke in the sense tunnel egresses unnecessarily
>> > drop harmless
>> > > > packets merely because they carry a "currently unused" (CU)
>> > > > combination of codepoints. This is similar to the problem
>> > ECN faced
>> > > > with firewalls dropping packets unless they were Not-ECT.
>> > > >
>> > > > (a) wasn't originally part of the changes I was
>> > proposing. But Anil
>> > > > Agawaal convinced me it is worth doing. Reasoning: If we
>> > are making a
>> > > > change to decap anyway, let's get it as right as possible. If we
>> > > > decide not to make any other changes to decap, I wouldn't
>> > fight for
>> > > > this one alone.
>> > > >
>> > > >
>> > > > >b) Inner: ECT(0), Outer: ECT(1)
>> > > > >         3168/4301: ECT(0)
>> > > > >         Change: ECT(1)
>> > > > >         Rationale: Treat ECT(1) as more severe congestion
>> > > > than ECT(0).
>> > > > >         My view: Longer discussion below.
>> > > > >
>> > > > >c) Inner: ECT(1), Outer: ECT(0)
>> > > > >         3168/4301: ECT(1)
>> > > > >         Change: ECT(1) (!!!)
>> > > > >         Rationale: "no currently legal transition" can
>> > > > produce this combo.
>> > > > >         My view: Not a good idea, among other reasons
>> > > > because it's at
>> > > > >                 odds with your "future standards action"
>> > > > explanation for
>> > > > >                 the first change.  I'd remove this change
>> > > > and instead deal
>> > > > >                 with the possibility of raising an alarm in
>> > > > a PCN-specific
>> > > > >                 draft rather than declaring this to be an
>> > > > alarm-able combo
>> > > > >                 in general.
>> > > > >
>> > > > >d) Inner: CE, Outer: ECT(1)
>> > > > >         3168/4301: CE
>> > > > >         Change: CE (!!!)
>> > > > >         Rationale & My view: <same as previous change>
>> > > >
>> > > > c) & d) aren't PCN specific.
>> > > >
>> > > > Reasoning for adding alarms (also said this eariler on list in
>> > > > conversation with Gorry):
>> > > > If security folks have concerns over possible attacks
>> > from "currently
>> > > > unused" combinations, I want to give them a way to manage the
>> > > > perceived threat without just dropping packets. RFC3168
>> > recommended
>> > > > alarms for CU combinations. In these cases, I believe
>> > RFC3168 simply
>> > > > overlooked that these are also CU combinations.
>> > > >
>> > > >
>> > > > >So, I would remove a) as not defensible (no good reason
>> > for change),
>> > > > >and I would push the discussion of adding alarms for c)
>> > and d) into
>> > > > >a PCN-specific document.
>> > > >
>> > > > c) & d) are not to do with PCN.
>> > > >
>> > > > My motivation is: While making change (b) to enable PCN, lets do
>> > > > (a,c,d) to make the IP layer as clean as possible, to
>> > ensure easier
>> > > > future evolvability.
>> > > >
>> > > > I don't see the issue. It says SHOULD log and MAY alarm. So the
>> > > > operator doesn't have to do anything.
>> > > >
>> > > > >The c) alarm in particular is only
>> > > > >meaningful if PCN or something like it is in use.
>> > > >
>> > > > No. The combination in c) always was CU, and it's still
>> > CU with PCN.
>> > > > None of the proposed changes are PCN-specific.
>> > > >
>> > > > Even (b) is intended to also be useful for any 3-level
>> > ECN while not
>> > > > precluding 2-level ECN and the nonce.
>> > > >
>> > > > NB:  3 level = 2 'congested' levels + 1 level not
>> > congested, e.g. VCP
>> > > > or PCN which were both mentioned in RFC4774. This doesn't
>> > count the
>> > > > not-ECN-capable state, which makes 4 states.
>> > > >
>> > > > >I don't view d)
>> > > > >as deserving of an alarm in general, as even though something has
>> > > > >happened, no damage has been done - this is in line with the
>> > > > >"be liberal in what you accept" principle of protocol design.
>> > > >
>> > > > That's why I'm proposing an optional alarm. Optional alarms *are*
>> > > > liberal, but they allow operators a way to address
>> > security concerns
>> > > > about 'known unknowns' [Rumsfelt], while still allowing new
>> > > > behaviours to be introduced.
>> > > >
>> > > >
>> > > > >That leaves b), which I think is the key desired change, because
>> > > > >it makes a third ECN code point usable through tunnels.
>> > > >
>> > > > Yes - exactly.
>> > > >
>> > > > >Much as
>> > > > >this might be useful for PCN, it may be too late to do this.  The
>> > > > >problem starts from the need to say that this draft
>> > updates RFC 4301.
>> > > > >The result will be a mix of 4301 and updated-4301 implementations
>> > > > >that render the third state difficult to use in practice.  The
>> > > > >problem's even worse for non-IPsec tunnels.
>> > > >
>> > > > The idea is that an operator wanting to implement PCN will have a
>> > > > spec they can point to in their invitation to tender. At the same
>> > > > time this spec both UPDATEs and is backward compatible
>> > with RFC3168 &
>> > > > RFC4301 because it only changes CU stuff at egress and
>> > the change at
>> > > > the ingress leads to the same output at the egress.
>> > Therefore vendors
>> > > > can update 3168 & 4301 tunnels lazily - only when their
>> > customers ask
>> > > > for the new features (whether for PCN or for something
>> > else like VCP).
>> > > >
>> > > > >I also read Appendix D
>> > > > >as saying that PCN is going the two-state route, not three.
>> > > >
>> > > > That's not how it should be read - I'll clarify it. The
>> > PCN w-g wants
>> > > > 3 states, but can't have them while ecn-tunnel is not proposed
>> > > > standard. Therefore, it has decided to define a 2-state
>> > baseline and
>> > > > a number of experimental schemes to work round this tunnel issue
>> > > > (plus a comparison doc about them all). It has agreed
>> > that eventually
>> > > > just one experimental scheme will be chosen for standards track:
>> > > > - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
>> > > > - 3-state (which uses an extra DSCP)
>> > > > - psdm (which requires out of band state)
>> > > > - sm (single marking gets 3 functions from 2 encoding states but
>> > > > admits admission control & termination aren't very precise in
>> > > > all cases)
>> > > >
>> > > > ___
>> > > > See part 2/2 for responses to your comments on specific parts
>> > > > of the test.
>> > > >
>> > > > Thanks again for the review
>> > > >
>> > > >
>> > > > Bob
>> > > >
>> > > >
>> > > >
>> > > > ________________________________________________________________
>> > > > Bob Briscoe,               Networks Research Centre, BT Research
>> > > >
>> > > >
>> > > >
>> >
>> > ________________________________________________________________
>> > Bob Briscoe,               Networks Research Centre, BT Research
>> >
>> >
>> >
>
>________________________________________________________________
>Bob Briscoe,               Networks Research Centre, BT Research

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From jmpolk@cisco.com  Mon Jul 20 18:37:56 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id C99C93A695F for <tsvwg@core3.amsl.com>; Mon, 20 Jul 2009 18:37:56 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -5.222
X-Spam-Level: 
X-Spam-Status: No, score=-5.222 tagged_above=-999 required=5 tests=[AWL=-0.623, BAYES_00=-2.599, DNS_FROM_OPENWHOIS=1.13, RCVD_IN_DNSWL_MED=-4, SARE_MLH_Stock1=0.87]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id DC8XFDcuwuum for <tsvwg@core3.amsl.com>; Mon, 20 Jul 2009 18:37:56 -0700 (PDT)
Received: from sj-iport-6.cisco.com (sj-iport-6.cisco.com [171.71.176.117]) by core3.amsl.com (Postfix) with ESMTP id 258C13A6DF5 for <tsvwg@ietf.org>; Mon, 20 Jul 2009 18:37:56 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AngHAG24ZEqrR7O6/2dsb2JhbACIQrEOiCOOdwWEDA
X-IronPort-AV: E=Sophos;i="4.43,237,1246838400"; d="scan'208";a="350532540"
Received: from sj-dkim-2.cisco.com ([171.71.179.186]) by sj-iport-6.cisco.com with ESMTP; 21 Jul 2009 01:37:56 +0000
Received: from sj-core-5.cisco.com (sj-core-5.cisco.com [171.71.177.238]) by sj-dkim-2.cisco.com (8.12.11/8.12.11) with ESMTP id n6L1bu2R024509 for <tsvwg@ietf.org>; Mon, 20 Jul 2009 18:37:56 -0700
Received: from xbh-sjc-211.amer.cisco.com (xbh-sjc-211.cisco.com [171.70.151.144]) by sj-core-5.cisco.com (8.13.8/8.14.3) with ESMTP id n6L1buME001492 for <tsvwg@ietf.org>; Tue, 21 Jul 2009 01:37:56 GMT
Received: from xfe-sjc-211.amer.cisco.com ([171.70.151.174]) by xbh-sjc-211.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 20 Jul 2009 18:37:56 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.4.23]) by xfe-sjc-211.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 20 Jul 2009 18:37:55 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 20 Jul 2009 20:37:54 -0500
To: tsvwg <tsvwg@ietf.org>
From: "James M. Polk" <jmpolk@cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-211oa2JB1nd00003a41@xfe-sjc-211.amer.cisco.com>
X-OriginalArrivalTime: 21 Jul 2009 01:37:56.0172 (UTC) FILETIME=[DF27B0C0:01CA09A3]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=295; t=1248140276; x=1249004276; c=relaxed/simple; s=sjdkim2002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20TSVWG's=20Agenda=20for=20Stockholm |Sender:=20; bh=I7Ta21llcr2I/0eF6JZMLJij3zNcAj1JFmss4bWNMYg=; b=fu7/o9aLsubNbbmruh9vaZ+ZdKvIy7MjRwGpvF+I8q/KI9r1Z+z1mA4pKV 06qhtTR0alYMZqGWLQdpWkDnN/NOAGKUIyrbY8CQmUhMkBgElMR/Tlkw2Asu rHI5P8MDX3;
Authentication-Results: sj-dkim-2; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim2002 verified; ); 
Subject: [tsvwg] TSVWG's Agenda for Stockholm
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 21 Jul 2009 01:37:56 -0000

TSVWG

Please review the WG agenda here

http://www.ietf.org/proceedings/75/agenda/tsvwg.txt

for any inconsistencies, or omissions.

Let the agenda bashing begin now.

TSVWG is meeting Thursday, July 30th, at 1300 local time in Stockholm.

See everyone there

James
TSVWG chair


From michawe@ulrik.uio.no  Tue Jul 21 02:54:56 2009
Return-Path: <michawe@ulrik.uio.no>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E21903A6880 for <tsvwg@core3.amsl.com>; Tue, 21 Jul 2009 02:54:56 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -5.514
X-Spam-Level: 
X-Spam-Status: No, score=-5.514 tagged_above=-999 required=5 tests=[AWL=0.215,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4, SARE_MLH_Stock1=0.87]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id IB3uMD9VoEid for <tsvwg@core3.amsl.com>; Tue, 21 Jul 2009 02:54:49 -0700 (PDT)
Received: from mail-out1.uio.no (mail-out1.uio.no [129.240.10.57]) by core3.amsl.com (Postfix) with ESMTP id B910C3A677D for <tsvwg@ietf.org>; Tue, 21 Jul 2009 02:54:49 -0700 (PDT)
Received: from mail-mx2.uio.no ([129.240.10.30]) by mail-out1.uio.no with esmtp (Exim 4.69) (envelope-from <michawe@ulrik.uio.no>) id 1MTBjs-0001Fu-Bh; Tue, 21 Jul 2009 11:34:24 +0200
Received: from w3prod-wm02.uio.no ([129.240.4.215] helo=webmail.uio.no) by mail-mx2.uio.no with esmtpsa (TLSv1:AES256-SHA:256) user michawe (Exim 4.69) (envelope-from <michawe@ulrik.uio.no>) id 1MTBjr-0003Sk-RL; Tue, 21 Jul 2009 11:34:24 +0200
Received: from 84.48.218.182 (SquirrelMail authenticated user michawe) by webmail.uio.no with HTTP; Tue, 21 Jul 2009 11:34:24 +0200
Message-ID: <a7e810bc55774e058c3d5be8381d6286.squirrel@webmail.uio.no>
In-Reply-To: <XFE-SJC-211oa2JB1nd00003a41@xfe-sjc-211.amer.cisco.com>
References: <XFE-SJC-211oa2JB1nd00003a41@xfe-sjc-211.amer.cisco.com>
Date: Tue, 21 Jul 2009 11:34:24 +0200
From: michawe@ifi.uio.no
To: "James M. Polk" <jmpolk@cisco.com>
User-Agent: SquirrelMail/1.4.19
MIME-Version: 1.0
Content-Type: text/plain;charset=iso-8859-1
Content-Transfer-Encoding: 8bit
X-UiO-Ratelimit-Test: rcpts/h 3 msgs/h 1 sum rcpts/h 3 sum msgs/h 1 total rcpts 2045 max rcpts/h 33 ratelimit 0
X-UiO-Spam-info: not spam, SpamAssassin (score=-5.0, required=5.0, autolearn=disabled, UIO_MAIL_IS_INTERNAL=-5, uiobl=NO, uiouri=NO)
X-UiO-Scanned: C6CE76391080DD05223902DF0C3311D8CDB1DDB2
X-UiO-SPAM-Test: remote_host: 129.240.4.215 spam_score: -49 maxlevel 80 minaction 2 bait 0 mail/h: 95 total 1513764 max/h 707 blacklist 0 greylist 0 ratelimit 0
Cc: tsvwg <tsvwg@ietf.org>, dragana.damjanovic@uibk.ac.at
Subject: Re: [tsvwg] TSVWG's Agenda for Stockholm
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 21 Jul 2009 09:54:57 -0000

Hi,

Michael - MulTFRC: TFRC with weighted fairness
draft-welzl-multfrc-00.txt

was omitted. Accident or decision? Please let me know.

Cheers,
Michael


> TSVWG
>
> Please review the WG agenda here
>
> http://www.ietf.org/proceedings/75/agenda/tsvwg.txt
>
> for any inconsistencies, or omissions.
>
> Let the agenda bashing begin now.
>
> TSVWG is meeting Thursday, July 30th, at 1300 local time in Stockholm.
>
> See everyone there
>
> James
> TSVWG chair
>
>



From rbriscoe@jungle.bt.co.uk  Tue Jul 21 10:35:48 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 6008D3A6782 for <tsvwg@core3.amsl.com>; Tue, 21 Jul 2009 10:35:48 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.279
X-Spam-Level: 
X-Spam-Status: No, score=-1.279 tagged_above=-999 required=5 tests=[AWL=-0.292, BAYES_00=-2.599, DNS_FROM_OPENWHOIS=1.13, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id LDQYnC4OKDCo for <tsvwg@core3.amsl.com>; Tue, 21 Jul 2009 10:35:41 -0700 (PDT)
Received: from smtp2.smtp.bt.com (smtp2.smtp.bt.com [217.32.164.150]) by core3.amsl.com (Postfix) with ESMTP id 435993A6978 for <tsvwg@ietf.org>; Tue, 21 Jul 2009 10:35:40 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp2.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Tue, 21 Jul 2009 18:35:27 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Tue, 21 Jul 2009 18:35:27 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248197727155; Tue, 21 Jul 2009 18:35:27 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6LHZNq9009754; Tue, 21 Jul 2009 18:35:23 +0100
Message-Id: <200907211735.n6LHZNq9009754@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Tue, 21 Jul 2009 18:35:22 +0100
To: "tsvwg IETF list" <tsvwg@ietf.org>, "PCN IETF list" <pcn@ietf.org>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 21 Jul 2009 17:35:27.0373 (UTC) FILETIME=[A2BD43D0:01CA0A29]
Subject: [tsvwg] draft-ietf-tsvwg-ecn-tunnel-03 update pending I-D queue
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 21 Jul 2009 17:35:48 -0000

tsvwg & pcn folks,

I've revised draft-ietf-tsvwg-ecn-tunnel to -03 after extensive 
review comments. Thanks for the deep reviews from Gorry Fairhurst, 
Michael Menth, Ingemar Johansson, Toby Moncaster and David Black.

I didn't upload it in time for the draft deadline (sorry), but I've 
uploaded -03 to my personal Web site, while we wait for the queue to re-open:
<http://www.bobbriscoe.net/pubs.html#ecn-tunnel>

I shall be presenting the draft in the Security Area Advisory Group 
next week, as well as in tsvwg.

As you may have noticed, there are still open discussions on the 
list, but I wanted to put the new revision up there, as it's had a 
major bashing. A summary of the changes is below...


Bob
===========================================================================

              Tunnelling of Explicit Congestion Notification
                      draft-ietf-tsvwg-ecn-tunnel-03

Abstract

    This document redefines how the explicit congestion notification
    (ECN) field of the IP header should be constructed on entry to and
    exit from any IP in IP tunnel.  On encapsulation it updates RFC3168
    to bring all IP in IP tunnels (v4 or v6) into line with RFC4301 IPsec
    ECN processing.  On decapsulation it updates both RFC3168 and RFC4301
    to add new behaviours for previously unused combinations of inner and
    outer header.  The new rules propagate the ECN field whether it is
    used to signal one or two severity levels of congestion, whereas
    before they propagated only one.  Tunnel endpoints can be updated in
    any order without affecting pre-existing uses of the ECN field
    (backward compatible).  Nonetheless, operators wanting to support two
    severity levels (e.g. for pre-congestion notification--PCN) can
    require compliance with this new specification.  A thorough analysis
    of the reasoning for these changes and the implications is included.


    From ietf-02 to ietf-03 (current):

       *  Functional changes:

          +  Corrected errors in recap of previous RFCs, which wrongly
             stated the different decapsulation behaviours of RFC3168 &
             RFC4301 with a Not-ECT inner header.  This also required
             corrections to the "Changes from Earlier RFCs" and the
             Motivations for these changes.

          +  Mandated that any future standards action SHOULD NOT use the
             ECT(0) codepoint as an indication of congestion, without
             giving strong reasons.

          +  Added optional alarm when decapsulating ECT(1) outer,
             ECT(0), but noted it would need to be disabled for
             2-severity level congestion (e.g.  PCN).

       *  Structural changes:

          +  Removed Document Roadmap which merely repeated the Contents
             (previously Section 1.2).

          +  Moved "Changes from Earlier RFCs" (Section 5) before
             Section 6 on Backward Compatibility and internally organised
             both by RFC, rather than by ingress then egress.

          +  Moved motivation for changing existing RFCs (Section 5.3) to
             after the changes are specified.

          +  Moved informative "Design Principles for Future Non-Default
             Schemes" after all the normative sections.

          +  Added Appendix A on early history of ECN tunnelling RFCs.

          +  Removed specialist appendix on "Relative Placement of
             Tunnelling and In-Path Load Regulation" (Appendix D in the
             -02 draft)

          +  Moved and updated specialist text on "Compromise on Decap
             with ECT(1) Inner and ECT(0) Outer" from Security
             Considerations to Appendix F

       *  Textual changes:

          +  Simplified vocabulary for non-native-english speakers

          +  Simplified Introduction and defined regularly used terms in
             an expanded Terminology section.

          +  More clearly distinguished statically configured tunnels
             from dynamic tunnel endpoint discovery, before explaining
             operating modes.

          +  Simplified, cut-down and clarified throughout

          +  Updated references.




________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From Black_David@emc.com  Thu Jul 23 14:44:48 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 7F08B3A6BA1 for <tsvwg@core3.amsl.com>; Thu, 23 Jul 2009 14:44:48 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.599
X-Spam-Level: 
X-Spam-Status: No, score=-7.599 tagged_above=-999 required=5 tests=[AWL=1.000,  BAYES_00=-2.599, GB_I_INVITATION=-2, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 4bzmrAAfEN8S for <tsvwg@core3.amsl.com>; Thu, 23 Jul 2009 14:44:46 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id 57C3F3A6A5F for <tsvwg@ietf.org>; Thu, 23 Jul 2009 14:44:46 -0700 (PDT)
Received: from hop04-l1d11-si03.isus.emc.com (HOP04-L1D11-SI03.isus.emc.com [10.254.111.23]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6NLiMKa003526 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 23 Jul 2009 17:44:22 -0400
Received: from mailhub.lss.emc.com (nagas.lss.emc.com [10.254.144.15]) by hop04-l1d11-si03.isus.emc.com (Tablus Interceptor); Thu, 23 Jul 2009 17:44:15 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2mp/Switch-3.3.2mp) with ESMTP id n6NLi5fq021089; Thu, 23 Jul 2009 17:44:05 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 23 Jul 2009 17:44:05 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Thu, 23 Jul 2009 17:44:04 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A034AA985@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
Thread-Index: AcoJMAAEREyypdcjQcyrRIJFNTjPRQCrMUjQ
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk> <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 23 Jul 2009 21:44:05.0489 (UTC) FILETIME=[B3749E10:01CA0BDE]
X-EMM-EM: Active
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 23 Jul 2009 21:44:48 -0000

Bob,

> RFC4301 seems to say: If the inner is Not-ECT, anything else in the=20
> outer must be an error. But whereas RFC3168 dropped the packet if=20
> there was CE in the outer, RFC4301 says forward it as Not-ECT=20
> in all cases.
>=20
> RFC4301 gives no justification for this rather important change from=20
> RFC3168. I've spent the last hour or so trawling the IPsec and tsvwg=20
> archives and can't find any discussion on this point either. Was it=20
> discussed anywhere?

Not that I recall.  I believe you're looking at a judgment call for
error case handling.  RFC 4301 was specified with close to the
minimum necessary to get ECN to work across an IPsec tunnel, as
security is an area where the default should be leaving out things
that aren't really necessary.  I believe that handling of this error
case was essentially treated as a "don't care."  Looking at your
alternative lines of thought:

> * If the ingress wrongly turned a Not-ECT inner into an ECT outer,=20
> then subsequently a congested router marked it CE, then it's unsafe=20
> to remove CE and forward Not-ECT - drop would have been safer.

The ingress is rather unlikely to have put ECT in the outer - that
clearly violates both RFC 3168 and RFC 4301, and is a bad idea.
Hence something's almost certainly broken ...

> * If something broken set CE on the outer even tho there was no=20
> congestion, it's safe.

If the packet shows up with Not-ECT in the inner and CE in the
outer, something's definitely broken, even if the breakage is
actually a bad Not-ECT -> ECT transition followed by a correct
downstream ECT->CE transition.  Not attempting to perform a
compensating fix is a reasonable course of action, IMHO ...
but I could also defend the 3168 drop behavior as within
reason, along the lines that you suggest.

Thanks,
--David
=20

> -----Original Message-----
> From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]=20
> Sent: Monday, July 20, 2009 7:48 AM
> To: Black, David
> Cc: tsvwg@ietf.org
> Subject: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
>=20
> David,
>=20
> I wanted to move on from optional alarms to my proposed functional=20
> changes, ie. your case a).
>=20
> But I have to admit a big error which makes that moot. In=20
> ecn-tunnel-02 I incorrectly reported what RFC3168 & RFC4301 said (I=20
> just re-read them). See the table below.
>=20
> IP in IP Decapsulation; Recap of Pre-existing Behaviour
> +------------------+----------------------------------------------+
> |  Incoming Inner  |             Incoming Outer Header            |
> |      Header      +---------+------------+------------+----------+
> |                  | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> +------------------+---------+------------+------------+----------+
> |     Not-ECT      | Not-ECT | drop       | drop       | drop=20
>     |<my error
> |     Not-ECT      | Not-ECT | Not-ECT    | Not-ECT    | drop=20
>     |<-3168
> |     Not-ECT      | Not-ECT | Not-ECT    | Not-ECT    |=20
> Not-ECT  |<-4301
> |      ECT(0)      |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
> |      ECT(1)      |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
> |        CE        |      CE |     CE     |     CE     |   CE     |
> +------------------+---------+------------+------------+----------+
>                     |                Outgoing Header               |
>                     +----------------------------------------------+
>=20
> RFC4301 seems to say: If the inner is Not-ECT, anything else in the=20
> outer must be an error. But whereas RFC3168 dropped the packet if=20
> there was CE in the outer, RFC4301 says forward it as Not-ECT=20
> in all cases.
>=20
> RFC4301 gives no justification for this rather important change from=20
> RFC3168. I've spent the last hour or so trawling the IPsec and tsvwg=20
> archives and can't find any discussion on this point either. Was it=20
> discussed anywhere?
>=20
> This RFC4301 behaviour seems unsafe, but perhaps understandable. It=20
> all depends how the situation arose:
> * If the ingress wrongly turned a Not-ECT inner into an ECT outer,=20
> then subsequently a congested router marked it CE, then it's unsafe=20
> to remove CE and forward Not-ECT - drop would have been safer.
> * If something broken set CE on the outer even tho there was no=20
> congestion, it's safe.
>=20
>=20
>=20
> Bob
>=20
> At 12:07 13/07/2009, Bob Briscoe wrote:
> >David,
> >
> >We're getting somewhere (if we set aside (a) for now).
> >
> > > At 17:09 10/07/2009, Black_David@emc.com wrote:
> >> > >Quick summary of the four changes proposed in the draft:
> >> > >
> >> > >a) Inner: Not-ECT, Outer: ECT(0)
> >> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> >> > >b) Inner: ECT(0), Outer: ECT(1)
> >> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
> >> > >c) Inner: ECT(1), Outer: ECT(0)
> >> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> >> > >d) Inner: CE, Outer: ECT(1)
> >> > >         RFC 3168/4301: CE, Change: CE(!!!)
> >> > >
> >> > >If we were designing from scratch, all of these changes would
> >> > >be fine things to do.  We're not, and especially after the
> >> > >discussion of how this impacts RFC 2003 gets added, this
> >> > >draft will be proposing changes to some of the most widely
> >> > >deployed mechanisms in the Internet.  To me, that strongly
> >> > >suggests a conservative design approach - focus on what really
> >> > >needs to be changed, and resist them temptation to fix other
> >> > >things that would be "nice to have".  In other words, IMHO
> >> > >the appropriate criteria for "fixing it" is a very strong
> >> > >cast that "it's broken".
> >> > >
> >> > >This is an area of principled disagreement on goals between
> >> > >Bob and myself that list discussion is unlikely to resolve.
> >> > >My view of the above four changes is that only b) makes it
> >> > >over the "really needs to be changed" bar (and even then,
> >> > >only with a *much* stronger statement of the PCN-based
> >> > >rationale than is in the current draft, and Bob concurs
> >> > >with strengthening the rationale).
> >> > >
> >> > >Turning to the other three changes, a) is clearly described
> >> > >as "nice to have", enough said.
> >> > >
> >> > >For c) and d), I don't believe the optional alarms are
> >> > >justified based on ECN (RFC 3168) alone.  A key observation
> >> > >is that this draft does not make the < b) ECT(0) [Inner]
> >> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> >> > >optional alarm.  What this means is that this draft not only
> >> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
> >> > >*assumes* that they are ordered in terms of priority,
> >> > >severity, or something analogous.  That is unwarranted as
> >> > >a general design assumption, IMHO.
> >> > >
> >> > >For this, reason, I think discussion of which transitions are
> >> > >optionally alarm-able in this fashion belongs in the context of
> >> > >a protocol that actually orders these codepoints, not in a
> >> > >general document that would serve to "just constrain future
> >> > >standards," borrowing Toby's words ;-).
> >> > >
> >> > >I could see discussion of optional alarms in an Appendix that
> >> > >discusses 3-valued PCN as an example, but not in the main body
> >> > >of the draft.  In the main body, it would be fine to say that
> >> > >specific uses of the ECN codepoints may assign meanings that
> >> > >cause some of these transitions to be appropriate for alarms
> >> > >(and point to the Appendix for a PCN example), but it would
> >> > >not be ok to constrain the assignment of those meanings by
> >> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> >> > >optionally alarm-able.
> >> > >
> >> > >Thanks,
> >> > >--David
> >> > >
> >> > >
> >> > > > -----Original Message-----
> >> > > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> >> > > > On Behalf Of Bob Briscoe
> >> > > > Sent: Thursday, July 09, 2009 3:08 PM
> >> > > > To: Black, David
> >> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> >> > > > Subject: Re: [tsvwg] review of
> >> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> >> > > >
> >> > > > David,
> >> > > >
> >> > > > Bringing to the list as agreed. I'll reply in 2=20
> parts. Part 1/2...
> >> > > >
> >> > > > Your comments get to the nub of the proposed egress
> >> > protocol changes.
> >> > > > Great... inline...
> >> > > >
> >> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> >> > > > >Bob,
> >> > > > >
> >> > > > >Sorry for the delay on this: things got entirely too
> >> > busy around the
> >> > > > >end of the quarter here.  Having re-read the draft from
> >> > scratch, I
> >> > > > >have a number of things to comment on ... I will be=20
> in Stockholm,
> >> > > > >and some of these items are going to need further discussion
> >> > > > >(private and/or public) in Stockholm.
> >> > > > >
> >> > > > >Starting with this table in Figure 4:
> >> > > > >
> >> > > >
> >> >=20
> >+------------------+----------------------------------------------+
> >> > > > >|  Incoming Inner  |             Incoming Outer Header
> >> >          |
> >> > > > >|      Header
> >> > +---------+------------+------------+----------+
> >> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
> >> >    CE    |
> >> > > >
> >> >=20
> >+------------------+---------+------------+------------+----------+
> >> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
> >> > drop(!!!)|
> >> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
> >> >   CE     |
> >> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
> >> >   CE     |
> >> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
> >> >   CE     |
> >> > > >
> >> >=20
> >+------------------+---------+------------+------------+----------+
> >> > > > >                    |                Outgoing Header
> >> >           |
> >> > > > >
> >> > +----------------------------------------------+
> >> > > > >
> >> > > > >My initial focus is on what needs to change from RFC
> >> > 3168/RFC 4301
> >> > > > >behavior and why.  Given that these two RFCs line up, I
> >> > think changes
> >> > > > >need strong motivation.  So, here are the four changes, and
> >> > > > my thoughts:
> >> > > > >
> >> > > > >a) Inner: Not-ECT, Outer: ECT(0)
> >> > > > >         3168/4301: drop (!!!)
> >> > > > >         Change: Not-ECT (!!!)
> >> > > > >         Rationale: "allow future standards actions=20
> to use this
> >> > > > >                 combination"
> >> > > > >         My view: This ain't broke, why are you fixing it?
> >> > > >
> >> > > > It is broke in the sense tunnel egresses unnecessarily
> >> > drop harmless
> >> > > > packets merely because they carry a "currently unused" (CU)
> >> > > > combination of codepoints. This is similar to the problem
> >> > ECN faced
> >> > > > with firewalls dropping packets unless they were Not-ECT.
> >> > > >
> >> > > > (a) wasn't originally part of the changes I was
> >> > proposing. But Anil
> >> > > > Agawaal convinced me it is worth doing. Reasoning: If we
> >> > are making a
> >> > > > change to decap anyway, let's get it as right as=20
> possible. If we
> >> > > > decide not to make any other changes to decap, I wouldn't
> >> > fight for
> >> > > > this one alone.
> >> > > >
> >> > > >
> >> > > > >b) Inner: ECT(0), Outer: ECT(1)
> >> > > > >         3168/4301: ECT(0)
> >> > > > >         Change: ECT(1)
> >> > > > >         Rationale: Treat ECT(1) as more severe congestion
> >> > > > than ECT(0).
> >> > > > >         My view: Longer discussion below.
> >> > > > >
> >> > > > >c) Inner: ECT(1), Outer: ECT(0)
> >> > > > >         3168/4301: ECT(1)
> >> > > > >         Change: ECT(1) (!!!)
> >> > > > >         Rationale: "no currently legal transition" can
> >> > > > produce this combo.
> >> > > > >         My view: Not a good idea, among other reasons
> >> > > > because it's at
> >> > > > >                 odds with your "future standards action"
> >> > > > explanation for
> >> > > > >                 the first change.  I'd remove this change
> >> > > > and instead deal
> >> > > > >                 with the possibility of raising an alarm in
> >> > > > a PCN-specific
> >> > > > >                 draft rather than declaring this to be an
> >> > > > alarm-able combo
> >> > > > >                 in general.
> >> > > > >
> >> > > > >d) Inner: CE, Outer: ECT(1)
> >> > > > >         3168/4301: CE
> >> > > > >         Change: CE (!!!)
> >> > > > >         Rationale & My view: <same as previous change>
> >> > > >
> >> > > > c) & d) aren't PCN specific.
> >> > > >
> >> > > > Reasoning for adding alarms (also said this eariler=20
> on list in
> >> > > > conversation with Gorry):
> >> > > > If security folks have concerns over possible attacks
> >> > from "currently
> >> > > > unused" combinations, I want to give them a way to manage the
> >> > > > perceived threat without just dropping packets. RFC3168
> >> > recommended
> >> > > > alarms for CU combinations. In these cases, I believe
> >> > RFC3168 simply
> >> > > > overlooked that these are also CU combinations.
> >> > > >
> >> > > >
> >> > > > >So, I would remove a) as not defensible (no good reason
> >> > for change),
> >> > > > >and I would push the discussion of adding alarms for c)
> >> > and d) into
> >> > > > >a PCN-specific document.
> >> > > >
> >> > > > c) & d) are not to do with PCN.
> >> > > >
> >> > > > My motivation is: While making change (b) to enable=20
> PCN, lets do
> >> > > > (a,c,d) to make the IP layer as clean as possible, to
> >> > ensure easier
> >> > > > future evolvability.
> >> > > >
> >> > > > I don't see the issue. It says SHOULD log and MAY=20
> alarm. So the
> >> > > > operator doesn't have to do anything.
> >> > > >
> >> > > > >The c) alarm in particular is only
> >> > > > >meaningful if PCN or something like it is in use.
> >> > > >
> >> > > > No. The combination in c) always was CU, and it's still
> >> > CU with PCN.
> >> > > > None of the proposed changes are PCN-specific.
> >> > > >
> >> > > > Even (b) is intended to also be useful for any 3-level
> >> > ECN while not
> >> > > > precluding 2-level ECN and the nonce.
> >> > > >
> >> > > > NB:  3 level =3D 2 'congested' levels + 1 level not
> >> > congested, e.g. VCP
> >> > > > or PCN which were both mentioned in RFC4774. This doesn't
> >> > count the
> >> > > > not-ECN-capable state, which makes 4 states.
> >> > > >
> >> > > > >I don't view d)
> >> > > > >as deserving of an alarm in general, as even though=20
> something has
> >> > > > >happened, no damage has been done - this is in line with the
> >> > > > >"be liberal in what you accept" principle of=20
> protocol design.
> >> > > >
> >> > > > That's why I'm proposing an optional alarm. Optional=20
> alarms *are*
> >> > > > liberal, but they allow operators a way to address
> >> > security concerns
> >> > > > about 'known unknowns' [Rumsfelt], while still allowing new
> >> > > > behaviours to be introduced.
> >> > > >
> >> > > >
> >> > > > >That leaves b), which I think is the key desired=20
> change, because
> >> > > > >it makes a third ECN code point usable through tunnels.
> >> > > >
> >> > > > Yes - exactly.
> >> > > >
> >> > > > >Much as
> >> > > > >this might be useful for PCN, it may be too late to=20
> do this.  The
> >> > > > >problem starts from the need to say that this draft
> >> > updates RFC 4301.
> >> > > > >The result will be a mix of 4301 and updated-4301=20
> implementations
> >> > > > >that render the third state difficult to use in=20
> practice.  The
> >> > > > >problem's even worse for non-IPsec tunnels.
> >> > > >
> >> > > > The idea is that an operator wanting to implement=20
> PCN will have a
> >> > > > spec they can point to in their invitation to=20
> tender. At the same
> >> > > > time this spec both UPDATEs and is backward compatible
> >> > with RFC3168 &
> >> > > > RFC4301 because it only changes CU stuff at egress and
> >> > the change at
> >> > > > the ingress leads to the same output at the egress.
> >> > Therefore vendors
> >> > > > can update 3168 & 4301 tunnels lazily - only when their
> >> > customers ask
> >> > > > for the new features (whether for PCN or for something
> >> > else like VCP).
> >> > > >
> >> > > > >I also read Appendix D
> >> > > > >as saying that PCN is going the two-state route, not three.
> >> > > >
> >> > > > That's not how it should be read - I'll clarify it. The
> >> > PCN w-g wants
> >> > > > 3 states, but can't have them while ecn-tunnel is=20
> not proposed
> >> > > > standard. Therefore, it has decided to define a 2-state
> >> > baseline and
> >> > > > a number of experimental schemes to work round this=20
> tunnel issue
> >> > > > (plus a comparison doc about them all). It has agreed
> >> > that eventually
> >> > > > just one experimental scheme will be chosen for=20
> standards track:
> >> > > > - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> >> > > > - 3-state (which uses an extra DSCP)
> >> > > > - psdm (which requires out of band state)
> >> > > > - sm (single marking gets 3 functions from 2=20
> encoding states but
> >> > > > admits admission control & termination aren't very precise in
> >> > > > all cases)
> >> > > >
> >> > > > ___
> >> > > > See part 2/2 for responses to your comments on specific parts
> >> > > > of the test.
> >> > > >
> >> > > > Thanks again for the review
> >> > > >
> >> > > >
> >> > > > Bob
> >> > > >
> >> > > >
> >> > > >
> >> > > >=20
> ________________________________________________________________
> >> > > > Bob Briscoe,               Networks Research Centre,=20
> BT Research
> >> > > >
> >> > > >
> >> > > >
> >> >
> >> > ________________________________________________________________
> >> > Bob Briscoe,               Networks Research Centre, BT Research
> >> >
> >> >
> >> >
> >
> >________________________________________________________________
> >Bob Briscoe,               Networks Research Centre, BT Research
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research=20
>=20
>=20
>=20

From Black_David@emc.com  Thu Jul 23 14:56:20 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 4CEE83A6A5F for <tsvwg@core3.amsl.com>; Thu, 23 Jul 2009 14:56:20 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.799
X-Spam-Level: 
X-Spam-Status: No, score=-6.799 tagged_above=-999 required=5 tests=[AWL=-0.200, BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id MfBZBOkKd9WH for <tsvwg@core3.amsl.com>; Thu, 23 Jul 2009 14:56:18 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id A1C7B3A68D1 for <tsvwg@ietf.org>; Thu, 23 Jul 2009 14:56:18 -0700 (PDT)
Received: from hop04-l1d11-si03.isus.emc.com (HOP04-L1D11-SI03.isus.emc.com [10.254.111.23]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6NLHLv0008845 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 23 Jul 2009 17:17:21 -0400
Received: from mailhub.lss.emc.com (nagas.lss.emc.com [10.254.144.15]) by hop04-l1d11-si03.isus.emc.com (Tablus Interceptor); Thu, 23 Jul 2009 17:17:17 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2mp/Switch-3.3.2mp) with ESMTP id n6NLHD8v023988; Thu, 23 Jul 2009 17:17:16 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 23 Jul 2009 17:16:17 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Thu, 23 Jul 2009 17:16:16 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A034AA976@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907201001.n6KA1Pha004613@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
Thread-Index: AcoJIRaqNlW4pEtbSaqDv6hLNLY+uACuIMAg
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp.emc.com> <200907131730.n6DHUWHa000642@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033673E9@CORPUSMX80A.corp.emc.com> <200907201001.n6KA1Pha004613@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 23 Jul 2009 21:16:17.0244 (UTC) FILETIME=[D11AA1C0:01CA0BDA]
X-EMM-EM: Active
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 23 Jul 2009 21:56:20 -0000

Bob,

Sorry for the delayed response.

> At least we now understand each other's position.
> We agree a set of alarms on all of b)c)d) would not be PCN-specific.
> And when looked at as a whole, missing out an alarm on b) but keeping=20
> alarms c)d) would be PCN-specific (or 2-severity-level-specific to be
precise).

Yes, that captures the important point of agreement on alarms.

> I am happy to say that all three of b)c)&d) SHOULD be logged and MAY=20
> be alarmed, then when I say that logging & alarms will need to be=20
> turned off if new standards actions use a combination, I can give PCN=20
> as an example where b) alarms should be turned off.
>=20
> As PCN is on the standards track, this mention of an exception for=20
> PCN ought to be in the body, not an appendix, IMO.
>=20
> Acceptable?

Yes, including discussion of PCN in the body.  My primary concern is
that the general framework not be specific to PCN-like requirements,
and this approach satisfies that concern.

Thanks,
--David
=20

> -----Original Message-----
> From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]=20
> Sent: Monday, July 20, 2009 6:01 AM
> To: Black, David
> Cc: tsvwg@ietf.org
> Subject: RE: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>=20
> David,
>=20
> At least we now understand each other's position.
> We agree a set of alarms on all of b)c)d) would not be PCN-specific.
> And when looked at as a whole, missing out an alarm on b) but keeping=20
> alarms c)d) would be PCN-specific (or=20
> 2-severity-level-specific to be precise).
>=20
> But when looking at each case on its individual merits, an optional=20
> alarm on c) or d) makes sense to pick up unexpected behaviours,=20
> whether or not b) is alarmed.
>=20
> I am happy to say that all three of b)c)&d) SHOULD be logged and MAY=20
> be alarmed, then when I say that logging & alarms will need to be=20
> turned off if new standards actions use a combination, I can give PCN=20
> as an example where b) alarms should be turned off.
>=20
> As PCN is on the standards track, this mention of an exception for=20
> PCN ought to be in the body, not an appendix, IMO.
>=20
> Acceptable?
>=20
> __________
> I have also decided to remove completely the alarm states '(!!!)'=20
> from my table recapping RFC3168 & RFC4301. Originally, I found the=20
> mention of logging in RFC3168, but on a re-read I shouldn't have been=20
> so certain of what it meant, as the text is rather vague. Also I've=20
> found there's actually no mention of alarms in RFC4301. So all we=20
> have is this sentence in RFC3168:
>=20
> "9.2.2.  Changes to the ECN Field within an IPsec Tunnel.
>=20
>     If the ECN Field is changed inappropriately within an=20
> IPsec tunnel,
>     and this change is detected at the tunnel egress, then=20
> the receipt of
>     a packet not satisfying the appropriate condition for its SA is an
>     auditable event. [...]"
>=20
> It doesn't actually enumerate the "inappropriate" cases, so an=20
> operator could construe this as already blessing alarms on b)c)d), at=20
> least in the IPsec case.
>=20
> [Sorry for spending a couple of days before responding...]
>=20
>=20
> Bob
>=20
> At 04:18 15/07/2009, Black_David@emc.com wrote:
> >Bob,
> >
> > > Optional alarms c) & d) have no more effect on PCN than=20
> on anything
> > > else, and they have no more impact on anything else whether or not
> > > PCN is there. They have absolutely nothing to do with PCN.
> >
> >I think you've missed the point; in isolation, the above statement
> >is correct, ... BUT ... the triple of <no b) alarm, c) alarm,
> >d) alarm> is PCN-specific, AFAICT ...
> >
> > > >c) Inner: ECT(1), Outer: ECT(0)
> > > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > >
> > > In RFC3168, if you get this combination, something along=20
> the tunnel
> > > has changed ECT(1) to ECT(0). That implies something weird has
> > > happened - not necessarily dangerous but, as no known or envisaged
> > > protocols do this, it could turn into something dangerous if the
> > > tunnel operator doesn't find out what's going on  - which needs an
> > > (optional) alarm to trigger some action.
> >
> >The problem with this reasoning is that wrt RFC3168, b) is the same
> >class of weirdness - ECT(0) has changed to ECT(1), yet no (optional)
> >alarm is proposed.  Keep in mind that ECT(0) and ECT (1) are *not*
> >ordered values in RFC3168 - they're unordered code points.  The
> >only escape I see is from this is that PCN is the "known or
> >envisaged" protocol that uses b) - is there something I've missed?
> >
> >The d) discussion has the same issue, hence it's snipped.
> >
> >[... snip ...]
> >
> > > Just because RFC3168 didn't say "MAY alarm" in these=20
> cases, doesn't
> > > mean it shouldn't have. I'm suggesting these were missed,=20
> that's all.
> >
> >And had RFC 3168 done that, it would also have suggested an optional
> >alarm for b).  If an optional alarm for b) is not wanted, the result
> >is PCN-specific and belongs in an Appendix, IMHO.  The result
> >enhances flexibility for future usage:
> >- Main body says that some unexpected changes within the tunnel MAY
> >         generate optional alarms, but expected ECN transitions (per
> >         RFC3168) SHOULD NOT generate alarms.
> >- Appendix discussing PCN as an example says that *if* ECT(0) and
> >         ECT(1) are ordered (true for PCN, false for RFC3168), *then*
> >         b) is expected behavior, but c) and d) deserve=20
> optional alarms.
> >
> >Thanks,
> >--David
> >
> >
> > > -----Original Message-----
> > > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > > Sent: Monday, July 13, 2009 1:30 PM
> > > To: Black, David
> > > Cc: tsvwg@ietf.org
> > > Subject: RE: [tsvwg] review of=20
> draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > >
> > > David,
> > >
> > > Optional alarms c) & d) have no more effect on PCN than=20
> on anything
> > > else, and they have no more impact on anything else whether or not
> > > PCN is there. They have absolutely nothing to do with PCN.
> > >
> > > >c) Inner: ECT(1), Outer: ECT(0)
> > > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > >
> > > In RFC3168, if you get this combination, something along=20
> the tunnel
> > > has changed ECT(1) to ECT(0). That implies something weird has
> > > happened - not necessarily dangerous but, as no known or envisaged
> > > protocols do this, it could turn into something dangerous if the
> > > tunnel operator doesn't find out what's going on  - which needs an
> > > (optional) alarm to trigger some action.
> > >
> > > >d) Inner: CE, Outer: ECT(1)
> > > >         RFC 3168/4301: CE, Change: CE(!!!)
> > >
> > > Again, RFC3168 full functionality mode says if the inner=20
> is CE, set
> > > the outer to ECT(0). So ECT(1) in the outer implies a similar
> > > weirdness has happened as above. So ditto: MAY alarm.
> > >
> > > Just because RFC3168 didn't say "MAY alarm" in these=20
> cases, doesn't
> > > mean it shouldn't have. I'm suggesting these were missed,=20
> that's all.
> > >
> > >
> > >
> > > Bob
> > >
> > > At 13:53 13/07/2009, Black_David@emc.com wrote:
> > > >Bob,
> > > >
> > > > > We're getting somewhere (if we set aside (a) for now).
> > > > >
> > > > > I'll first try to reverse engineer how the way we=20
> think differs...
> > > > > - You seem to be thinking of different ECN tunnel=20
> behaviours for
> > > > > different types of tunnel (IPsec, non-IPsec, PCN,=20
> non-PCN etc).
> > > > > - My goal with this draft is to find a lowest common
> > > denominator ECN
> > > > > tunnel behaviour for all types of tunnel (IPsec,
> > > non-IPsec, 3-level,
> > > > > 2-level, ECN not supported, whatever).
> > > >
> > > >We're actually aligned on this - I agree with you that=20
> the latter is
> > > >the goal.  However, we disagree on what the lowest=20
> common denominator
> > > >is ...
> > > >
> > > >[... snip ...]
> > > >
> > > > > Because I'm trying to find the lowest common denominator,
> > > I say raise
> > > > > alarms for c) & d) whether using PCN or not.
> > > >
> > > >If PCN is not in use, alarms on c) and d) are not needed, IMHO
> > > >(e.g., cf. RFC 3168).
> > > >
> > > >I believe no alarms on c) and d) is the lowest common=20
> denominator.
> > > >What this means is that the baseline (lowest common denominator)
> > > >alarms only when a not-ECT Inner packet comes through with an ECN
> > > >codepoint in Outer header.
> > > >
> > > >I'd state that the baseline is no alarms added, and then=20
> PCN-specific
> > > >text can then explain why for PCN:
> > > >- c) and d) are alarming, but
> > > >- b) is expected behavior.
> > > >
> > > > > Then the question is, "Should we alarm on b) if PCN is _not_
> > > > > being used?"
> > > > >
> > > > > Is this what we really disagree on?
> > > >
> > > >No - IMHO, we should not alarm on b), c) or d) unless something
> > > >like PCN is being used.
> > > >
> > > >Thanks,
> > > >--David
> > > >
> > > > > -----Original Message-----
> > > > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > > > > On Behalf Of Bob Briscoe
> > > > > Sent: Monday, July 13, 2009 7:07 AM
> > > > > To: Black, David
> > > > > Cc: tsvwg@ietf.org
> > > > > Subject: Re: [tsvwg] review of
> > > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > > > >
> > > > > David,
> > > > >
> > > > > We're getting somewhere (if we set aside (a) for now).
> > > > >
> > > > > I'll first try to reverse engineer how the way we=20
> think differs...
> > > > > - You seem to be thinking of different ECN tunnel=20
> behaviours for
> > > > > different types of tunnel (IPsec, non-IPsec, PCN,=20
> non-PCN etc).
> > > > > - My goal with this draft is to find a lowest common
> > > denominator ECN
> > > > > tunnel behaviour for all types of tunnel (IPsec,
> > > non-IPsec, 3-level,
> > > > > 2-level, ECN not supported, whatever).
> > > > >
> > > > > This separation is important in both directions:
> > > > > - all nodes should be able to expect the ECN field to
> > > behave in the
> > > > > same way whether or not tunnels intervene without=20
> their control
> > > > > - all tunnels should be able to process the ECN field=20
> in the same
> > > > > way, whatever ECN is being used for, because tunnels=20
> get deployed
> > > > > independently of congestion control schemes like ECN & PCN.
> > > > >
> > > > > However, at encap I've had to accept that just one mode is
> > > > > impossible, as we still need the two RFC3168=20
> behaviours for safety
> > > > > with legacy egresses.
> > > > >
> > > > > now inline...
> > > > >
> > > > > At 23:28 10/07/2009, Black_David@emc.com wrote:
> > > > > >Bob,
> > > > > >
> > > > > >Actually it's only about a) if you can accept the following
> > > > > >approach for the optional alarms in c) and d):
> > > > > >
> > > > > >- Main body of draft says that specific uses of ECN may raise
> > > > > >         optional decapsulation alarms for selected
> > > cases in which
> > > > > >         the inner and outer header values don't match.
> > > > >
> > > > > Well, it's not that they don't match, it's that the
> > > combination isn't
> > > > > expected to arise - it's currently unused. But I know
> > > what you meant.
> > > > >
> > > > > >- An Appendix that discusses PCN as an example=20
> points out that
> > > > > >         for PCN, it is appropriate to raise alarms for
> > > c) and d).
> > > > >
> > > > > I think what you mean is: if PCN weren't being used, it
> > > would be _as_
> > > > > appropriate to raise an alarm for b) as it is for c) & d).
> > > > >
> > > > >
> > > > > >As for "How does this have anything to do with=20
> PCN?", the answer
> > > > > >to that question is the same as the answer to "Why does the
> > > > > >draft not propose an optional alarm for case b)?" ...  the
> > > > > >answer is probably some form of "no current use", in=20
> which case,
> > > > > >what's the current use for the c) and d) alarms? ...=20
> it's PCN.
> > > > >
> > > > > Because I'm trying to find the lowest common denominator,
> > > I say raise
> > > > > alarms for c) & d) whether using PCN or not.
> > > > >
> > > > > Then the question is, "Should we alarm on b) if PCN is _not_
> > > > > being used?"
> > > > >
> > > > > Is this what we really disagree on? If so, we could say:
> > > > > "MAY alarm on c) & d). MAY also alarm on b) if certain that no
> > > > > 3-level scheme is being used through the tunnel, but NOT
> > > RECOMMENDED
> > > > > due to the extra complexity of finding out whether a 3-level
> > > > > scheme is in use."
> > > > >
> > > > > Does this help? Or have I still mis-estimated where you're
> > > > > coming from?
> > > > >
> > > > >
> > > > > Bob
> > > > >
> > > >
> > > >[... snip ...]
> > >
> > > ________________________________________________________________
> > > Bob Briscoe,               Networks Research Centre, BT Research
> > >
> > >
> > >
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research=20
>=20
>=20
>=20

From rbriscoe@jungle.bt.co.uk  Fri Jul 24 01:51:25 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id AAB123A68F3 for <tsvwg@core3.amsl.com>; Fri, 24 Jul 2009 01:51:25 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.803
X-Spam-Level: 
X-Spam-Status: No, score=-2.803 tagged_above=-999 required=5 tests=[AWL=1.314,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id IS0Z7N04-5ix for <tsvwg@core3.amsl.com>; Fri, 24 Jul 2009 01:51:19 -0700 (PDT)
Received: from smtp4.smtp.bt.com (smtp4.smtp.bt.com [217.32.164.151]) by core3.amsl.com (Postfix) with ESMTP id 22A573A62C1 for <tsvwg@ietf.org>; Fri, 24 Jul 2009 01:51:18 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp4.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 24 Jul 2009 09:49:48 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Fri, 24 Jul 2009 09:49:48 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248425387421; Fri, 24 Jul 2009 09:49:47 +0100
Received: from mut.jungle.bt.co.uk ([10.215.130.87]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6O8nctP026517; Fri, 24 Jul 2009 09:49:38 +0100
Message-Id: <200907240849.n6O8nctP026517@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Fri, 24 Jul 2009 09:49:29 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A034AA985@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk> <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AA985@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 24 Jul 2009 08:49:48.0291 (UTC) FILETIME=[B3388530:01CA0C3B]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 24 Jul 2009 08:51:25 -0000

David,

Thanks for the insight into the likely reasoning.

 From a transport perspective, I think now is the right time to say 
safety needs greater attention...
- there's a desire to change 4301 decap anyway to support 2 severity levels,
- safety's become more important since 4301 - greater chance of errors...

Guys I've talked to in the pseudowire area are considering using ECN 
for middlebox to middlebox congestion control (a bit like PCN). 
Pseudowires (or PCN) might or might not also require IPsec. But it's 
quite an art to turn on ECN across a tunnelled segment and do no harm 
in both the following cases:
- if e2e ECN is on, avoiding polluting e2e ECN signals
- if e2e ECN is off, avoiding turning it on
(see recent PCN list).

This all means the chances that an error will occur increases, either 
in broken implementations or broken deployments.

The new rules in ecn-tunnel are designed so that tunnels should "just 
work" and "safely", even if they are deployed by someone who doesn't 
know they have to cater for an in-path internal ECN loop as well.

Q. What do you think of extending drop to both CE and ECT(1) outer if 
there's a Not-ECT inner (see table)?
Reasoning: As these are both currently error conditions and likely to 
remain so, rather than two different safety rules for different 
circumstances just make tunnels safe in all potential cases. 
Otherwise someone has to remember to configure a different safety 
rule on all their tunnels when PCN or pseudowires get deployed in 
their network.

+----------------+----------------------------------------------+
| Incoming Inner |             Incoming Outer Header            |
|     Header     +---------+------------+------------+----------+
|                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
+----------------+---------+------------+------------+----------+
|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    | drop     |<-3168
|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |Not-ECT   |<-4301
|    Not-ECT     | Not-ECT | Not-ECT    | drop       | drop     |<-ecn-tun
+----------------+---------+------------+------------+----------+
                  |                Outgoing Header               |
                  +----------------------------------------------+


Bob

At 22:44 23/07/2009, Black_David@emc.com wrote:
>Bob,
>
> > RFC4301 seems to say: If the inner is Not-ECT, anything else in the
> > outer must be an error. But whereas RFC3168 dropped the packet if
> > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > in all cases.
> >
> > RFC4301 gives no justification for this rather important change from
> > RFC3168. I've spent the last hour or so trawling the IPsec and tsvwg
> > archives and can't find any discussion on this point either. Was it
> > discussed anywhere?
>
>Not that I recall.  I believe you're looking at a judgment call for
>error case handling.  RFC 4301 was specified with close to the
>minimum necessary to get ECN to work across an IPsec tunnel, as
>security is an area where the default should be leaving out things
>that aren't really necessary.  I believe that handling of this error
>case was essentially treated as a "don't care."  Looking at your
>alternative lines of thought:
>
> > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > then subsequently a congested router marked it CE, then it's unsafe
> > to remove CE and forward Not-ECT - drop would have been safer.
>
>The ingress is rather unlikely to have put ECT in the outer - that
>clearly violates both RFC 3168 and RFC 4301, and is a bad idea.
>Hence something's almost certainly broken ...
>
> > * If something broken set CE on the outer even tho there was no
> > congestion, it's safe.
>
>If the packet shows up with Not-ECT in the inner and CE in the
>outer, something's definitely broken, even if the breakage is
>actually a bad Not-ECT -> ECT transition followed by a correct
>downstream ECT->CE transition.  Not attempting to perform a
>compensating fix is a reasonable course of action, IMHO ...
>but I could also defend the 3168 drop behavior as within
>reason, along the lines that you suggest.
>
>Thanks,
>--David
>
>
> > -----Original Message-----
> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > Sent: Monday, July 20, 2009 7:48 AM
> > To: Black, David
> > Cc: tsvwg@ietf.org
> > Subject: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
> >
> > David,
> >
> > I wanted to move on from optional alarms to my proposed functional
> > changes, ie. your case a).
> >
> > But I have to admit a big error which makes that moot. In
> > ecn-tunnel-02 I incorrectly reported what RFC3168 & RFC4301 said (I
> > just re-read them). See the table below.
> >
> > IP in IP Decapsulation; Recap of Pre-existing Behaviour
> >+----------------+----------------------------------------------+
> >| Incoming Inner |             Incoming Outer Header            |
> >|     Header     +---------+------------+------------+----------+
> >|                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> >+----------------+---------+------------+------------+----------+
> >|    Not-ECT     | Not-ECT | drop       | drop       | drop     |<my error
> >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    | drop     |<-3168
> >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |Not-ECT   |<-4301
> >|     ECT(0)     |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
> >|     ECT(1)     |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
> >|       CE       |      CE |     CE     |     CE     |   CE     |
> >+----------------+---------+------------+------------+----------+
> >                 |                Outgoing Header               |
> >                 +----------------------------------------------+
> >
> > RFC4301 seems to say: If the inner is Not-ECT, anything else in the
> > outer must be an error. But whereas RFC3168 dropped the packet if
> > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > in all cases.
> >
> > RFC4301 gives no justification for this rather important change from
> > RFC3168. I've spent the last hour or so trawling the IPsec and tsvwg
> > archives and can't find any discussion on this point either. Was it
> > discussed anywhere?
> >
> > This RFC4301 behaviour seems unsafe, but perhaps understandable. It
> > all depends how the situation arose:
> > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > then subsequently a congested router marked it CE, then it's unsafe
> > to remove CE and forward Not-ECT - drop would have been safer.
> > * If something broken set CE on the outer even tho there was no
> > congestion, it's safe.
> >
> >
> >
> > Bob
> >
> > At 12:07 13/07/2009, Bob Briscoe wrote:
> > >David,
> > >
> > >We're getting somewhere (if we set aside (a) for now).
> > >
> > > > At 17:09 10/07/2009, Black_David@emc.com wrote:
> > >> > >Quick summary of the four changes proposed in the draft:
> > >> > >
> > >> > >a) Inner: Not-ECT, Outer: ECT(0)
> > >> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> > >> > >b) Inner: ECT(0), Outer: ECT(1)
> > >> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
> > >> > >c) Inner: ECT(1), Outer: ECT(0)
> > >> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > >> > >d) Inner: CE, Outer: ECT(1)
> > >> > >         RFC 3168/4301: CE, Change: CE(!!!)
> > >> > >
> > >> > >If we were designing from scratch, all of these changes would
> > >> > >be fine things to do.  We're not, and especially after the
> > >> > >discussion of how this impacts RFC 2003 gets added, this
> > >> > >draft will be proposing changes to some of the most widely
> > >> > >deployed mechanisms in the Internet.  To me, that strongly
> > >> > >suggests a conservative design approach - focus on what really
> > >> > >needs to be changed, and resist them temptation to fix other
> > >> > >things that would be "nice to have".  In other words, IMHO
> > >> > >the appropriate criteria for "fixing it" is a very strong
> > >> > >cast that "it's broken".
> > >> > >
> > >> > >This is an area of principled disagreement on goals between
> > >> > >Bob and myself that list discussion is unlikely to resolve.
> > >> > >My view of the above four changes is that only b) makes it
> > >> > >over the "really needs to be changed" bar (and even then,
> > >> > >only with a *much* stronger statement of the PCN-based
> > >> > >rationale than is in the current draft, and Bob concurs
> > >> > >with strengthening the rationale).
> > >> > >
> > >> > >Turning to the other three changes, a) is clearly described
> > >> > >as "nice to have", enough said.
> > >> > >
> > >> > >For c) and d), I don't believe the optional alarms are
> > >> > >justified based on ECN (RFC 3168) alone.  A key observation
> > >> > >is that this draft does not make the < b) ECT(0) [Inner]
> > >> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> > >> > >optional alarm.  What this means is that this draft not only
> > >> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
> > >> > >*assumes* that they are ordered in terms of priority,
> > >> > >severity, or something analogous.  That is unwarranted as
> > >> > >a general design assumption, IMHO.
> > >> > >
> > >> > >For this, reason, I think discussion of which transitions are
> > >> > >optionally alarm-able in this fashion belongs in the context of
> > >> > >a protocol that actually orders these codepoints, not in a
> > >> > >general document that would serve to "just constrain future
> > >> > >standards," borrowing Toby's words ;-).
> > >> > >
> > >> > >I could see discussion of optional alarms in an Appendix that
> > >> > >discusses 3-valued PCN as an example, but not in the main body
> > >> > >of the draft.  In the main body, it would be fine to say that
> > >> > >specific uses of the ECN codepoints may assign meanings that
> > >> > >cause some of these transitions to be appropriate for alarms
> > >> > >(and point to the Appendix for a PCN example), but it would
> > >> > >not be ok to constrain the assignment of those meanings by
> > >> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> > >> > >optionally alarm-able.
> > >> > >
> > >> > >Thanks,
> > >> > >--David
> > >> > >
> > >> > >
> > >> > > > -----Original Message-----
> > >> > > > From: tsvwg-bounces@ietf.org [mailto:tsvwg-bounces@ietf.org]
> > >> > > > On Behalf Of Bob Briscoe
> > >> > > > Sent: Thursday, July 09, 2009 3:08 PM
> > >> > > > To: Black, David
> > >> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > >> > > > Subject: Re: [tsvwg] review of
> > >> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > >> > > >
> > >> > > > David,
> > >> > > >
> > >> > > > Bringing to the list as agreed. I'll reply in 2
> > parts. Part 1/2...
> > >> > > >
> > >> > > > Your comments get to the nub of the proposed egress
> > >> > protocol changes.
> > >> > > > Great... inline...
> > >> > > >
> > >> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > >> > > > >Bob,
> > >> > > > >
> > >> > > > >Sorry for the delay on this: things got entirely too
> > >> > busy around the
> > >> > > > >end of the quarter here.  Having re-read the draft from
> > >> > scratch, I
> > >> > > > >have a number of things to comment on ... I will be
> > in Stockholm,
> > >> > > > >and some of these items are going to need further discussion
> > >> > > > >(private and/or public) in Stockholm.
> > >> > > > >
> > >> > > > >Starting with this table in Figure 4:
> > >> > > > >
> > >> > > >
> > >> >
> > >+------------------+----------------------------------------------+
> > >> > > > >|  Incoming Inner  |             Incoming Outer Header
> > >> >          |
> > >> > > > >|      Header
> > >> > +---------+------------+------------+----------+
> > >> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
> > >> >    CE    |
> > >> > > >
> > >> >
> > >+------------------+---------+------------+------------+----------+
> > >> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
> > >> > drop(!!!)|
> > >> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
> > >> >   CE     |
> > >> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
> > >> >   CE     |
> > >> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
> > >> >   CE     |
> > >> > > >
> > >> >
> > >+------------------+---------+------------+------------+----------+
> > >> > > > >                    |                Outgoing Header
> > >> >           |
> > >> > > > >
> > >> > +----------------------------------------------+
> > >> > > > >
> > >> > > > >My initial focus is on what needs to change from RFC
> > >> > 3168/RFC 4301
> > >> > > > >behavior and why.  Given that these two RFCs line up, I
> > >> > think changes
> > >> > > > >need strong motivation.  So, here are the four changes, and
> > >> > > > my thoughts:
> > >> > > > >
> > >> > > > >a) Inner: Not-ECT, Outer: ECT(0)
> > >> > > > >         3168/4301: drop (!!!)
> > >> > > > >         Change: Not-ECT (!!!)
> > >> > > > >         Rationale: "allow future standards actions
> > to use this
> > >> > > > >                 combination"
> > >> > > > >         My view: This ain't broke, why are you fixing it?
> > >> > > >
> > >> > > > It is broke in the sense tunnel egresses unnecessarily
> > >> > drop harmless
> > >> > > > packets merely because they carry a "currently unused" (CU)
> > >> > > > combination of codepoints. This is similar to the problem
> > >> > ECN faced
> > >> > > > with firewalls dropping packets unless they were Not-ECT.
> > >> > > >
> > >> > > > (a) wasn't originally part of the changes I was
> > >> > proposing. But Anil
> > >> > > > Agawaal convinced me it is worth doing. Reasoning: If we
> > >> > are making a
> > >> > > > change to decap anyway, let's get it as right as
> > possible. If we
> > >> > > > decide not to make any other changes to decap, I wouldn't
> > >> > fight for
> > >> > > > this one alone.
> > >> > > >
> > >> > > >
> > >> > > > >b) Inner: ECT(0), Outer: ECT(1)
> > >> > > > >         3168/4301: ECT(0)
> > >> > > > >         Change: ECT(1)
> > >> > > > >         Rationale: Treat ECT(1) as more severe congestion
> > >> > > > than ECT(0).
> > >> > > > >         My view: Longer discussion below.
> > >> > > > >
> > >> > > > >c) Inner: ECT(1), Outer: ECT(0)
> > >> > > > >         3168/4301: ECT(1)
> > >> > > > >         Change: ECT(1) (!!!)
> > >> > > > >         Rationale: "no currently legal transition" can
> > >> > > > produce this combo.
> > >> > > > >         My view: Not a good idea, among other reasons
> > >> > > > because it's at
> > >> > > > >                 odds with your "future standards action"
> > >> > > > explanation for
> > >> > > > >                 the first change.  I'd remove this change
> > >> > > > and instead deal
> > >> > > > >                 with the possibility of raising an alarm in
> > >> > > > a PCN-specific
> > >> > > > >                 draft rather than declaring this to be an
> > >> > > > alarm-able combo
> > >> > > > >                 in general.
> > >> > > > >
> > >> > > > >d) Inner: CE, Outer: ECT(1)
> > >> > > > >         3168/4301: CE
> > >> > > > >         Change: CE (!!!)
> > >> > > > >         Rationale & My view: <same as previous change>
> > >> > > >
> > >> > > > c) & d) aren't PCN specific.
> > >> > > >
> > >> > > > Reasoning for adding alarms (also said this eariler
> > on list in
> > >> > > > conversation with Gorry):
> > >> > > > If security folks have concerns over possible attacks
> > >> > from "currently
> > >> > > > unused" combinations, I want to give them a way to manage the
> > >> > > > perceived threat without just dropping packets. RFC3168
> > >> > recommended
> > >> > > > alarms for CU combinations. In these cases, I believe
> > >> > RFC3168 simply
> > >> > > > overlooked that these are also CU combinations.
> > >> > > >
> > >> > > >
> > >> > > > >So, I would remove a) as not defensible (no good reason
> > >> > for change),
> > >> > > > >and I would push the discussion of adding alarms for c)
> > >> > and d) into
> > >> > > > >a PCN-specific document.
> > >> > > >
> > >> > > > c) & d) are not to do with PCN.
> > >> > > >
> > >> > > > My motivation is: While making change (b) to enable
> > PCN, lets do
> > >> > > > (a,c,d) to make the IP layer as clean as possible, to
> > >> > ensure easier
> > >> > > > future evolvability.
> > >> > > >
> > >> > > > I don't see the issue. It says SHOULD log and MAY
> > alarm. So the
> > >> > > > operator doesn't have to do anything.
> > >> > > >
> > >> > > > >The c) alarm in particular is only
> > >> > > > >meaningful if PCN or something like it is in use.
> > >> > > >
> > >> > > > No. The combination in c) always was CU, and it's still
> > >> > CU with PCN.
> > >> > > > None of the proposed changes are PCN-specific.
> > >> > > >
> > >> > > > Even (b) is intended to also be useful for any 3-level
> > >> > ECN while not
> > >> > > > precluding 2-level ECN and the nonce.
> > >> > > >
> > >> > > > NB:  3 level = 2 'congested' levels + 1 level not
> > >> > congested, e.g. VCP
> > >> > > > or PCN which were both mentioned in RFC4774. This doesn't
> > >> > count the
> > >> > > > not-ECN-capable state, which makes 4 states.
> > >> > > >
> > >> > > > >I don't view d)
> > >> > > > >as deserving of an alarm in general, as even though
> > something has
> > >> > > > >happened, no damage has been done - this is in line with the
> > >> > > > >"be liberal in what you accept" principle of
> > protocol design.
> > >> > > >
> > >> > > > That's why I'm proposing an optional alarm. Optional
> > alarms *are*
> > >> > > > liberal, but they allow operators a way to address
> > >> > security concerns
> > >> > > > about 'known unknowns' [Rumsfelt], while still allowing new
> > >> > > > behaviours to be introduced.
> > >> > > >
> > >> > > >
> > >> > > > >That leaves b), which I think is the key desired
> > change, because
> > >> > > > >it makes a third ECN code point usable through tunnels.
> > >> > > >
> > >> > > > Yes - exactly.
> > >> > > >
> > >> > > > >Much as
> > >> > > > >this might be useful for PCN, it may be too late to
> > do this.  The
> > >> > > > >problem starts from the need to say that this draft
> > >> > updates RFC 4301.
> > >> > > > >The result will be a mix of 4301 and updated-4301
> > implementations
> > >> > > > >that render the third state difficult to use in
> > practice.  The
> > >> > > > >problem's even worse for non-IPsec tunnels.
> > >> > > >
> > >> > > > The idea is that an operator wanting to implement
> > PCN will have a
> > >> > > > spec they can point to in their invitation to
> > tender. At the same
> > >> > > > time this spec both UPDATEs and is backward compatible
> > >> > with RFC3168 &
> > >> > > > RFC4301 because it only changes CU stuff at egress and
> > >> > the change at
> > >> > > > the ingress leads to the same output at the egress.
> > >> > Therefore vendors
> > >> > > > can update 3168 & 4301 tunnels lazily - only when their
> > >> > customers ask
> > >> > > > for the new features (whether for PCN or for something
> > >> > else like VCP).
> > >> > > >
> > >> > > > >I also read Appendix D
> > >> > > > >as saying that PCN is going the two-state route, not three.
> > >> > > >
> > >> > > > That's not how it should be read - I'll clarify it. The
> > >> > PCN w-g wants
> > >> > > > 3 states, but can't have them while ecn-tunnel is
> > not proposed
> > >> > > > standard. Therefore, it has decided to define a 2-state
> > >> > baseline and
> > >> > > > a number of experimental schemes to work round this
> > tunnel issue
> > >> > > > (plus a comparison doc about them all). It has agreed
> > >> > that eventually
> > >> > > > just one experimental scheme will be chosen for
> > standards track:
> > >> > > > - 3-in-1 (the simplest, which depends on (b) of ecn-tunnel)
> > >> > > > - 3-state (which uses an extra DSCP)
> > >> > > > - psdm (which requires out of band state)
> > >> > > > - sm (single marking gets 3 functions from 2
> > encoding states but
> > >> > > > admits admission control & termination aren't very precise in
> > >> > > > all cases)
> > >> > > >
> > >> > > > ___
> > >> > > > See part 2/2 for responses to your comments on specific parts
> > >> > > > of the test.
> > >> > > >
> > >> > > > Thanks again for the review
> > >> > > >
> > >> > > >
> > >> > > > Bob
> > >> > > >
> > >> > > >
> > >> > > >
> > >> > > >
> > ________________________________________________________________
> > >> > > > Bob Briscoe,               Networks Research Centre,
> > BT Research
> > >> > > >
> > >> > > >
> > >> > > >
> > >> >
> > >> > ________________________________________________________________
> > >> > Bob Briscoe,               Networks Research Centre, BT Research
> > >> >
> > >> >
> > >> >
> > >
> > >________________________________________________________________
> > >Bob Briscoe,               Networks Research Centre, BT Research
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From rbriscoe@jungle.bt.co.uk  Fri Jul 24 15:42:23 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 1C0E23A679C for <tsvwg@core3.amsl.com>; Fri, 24 Jul 2009 15:42:23 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.891
X-Spam-Level: 
X-Spam-Status: No, score=-1.891 tagged_above=-999 required=5 tests=[AWL=0.226,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id DKZP934hRAVK for <tsvwg@core3.amsl.com>; Fri, 24 Jul 2009 15:42:17 -0700 (PDT)
Received: from smtp4.smtp.bt.com (smtp4.smtp.bt.com [217.32.164.151]) by core3.amsl.com (Postfix) with ESMTP id 15EBE3A68B4 for <tsvwg@ietf.org>; Fri, 24 Jul 2009 15:42:16 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp4.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Fri, 24 Jul 2009 23:42:16 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Fri, 24 Jul 2009 23:42:16 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248475335199; Fri, 24 Jul 2009 23:42:15 +0100
Received: from mut.jungle.bt.co.uk ([10.73.64.34]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6OMgC3b012036; Fri, 24 Jul 2009 23:42:13 +0100
Message-Id: <200907242242.n6OMgC3b012036@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Fri, 24 Jul 2009 23:41:23 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A034AA976@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <200907131107.n6DB74cn016300@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033668D6@CORPUSMX80A.corp.emc.com> <200907131730.n6DHUWHa000642@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A033673E9@CORPUSMX80A.corp.emc.com> <200907201001.n6KA1Pha004613@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AA976@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 24 Jul 2009 22:42:16.0486 (UTC) FILETIME=[FEAA3C60:01CA0CAF]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] review of draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 24 Jul 2009 22:42:23 -0000

David,

At 22:16 23/07/2009, Black_David@emc.com wrote:
> > Acceptable?
>
>Yes, including discussion of PCN in the body.  My primary concern is
>that the general framework not be specific to PCN-like requirements,
>and this approach satisfies that concern.

Good.

Here's my logic for why I concede you could be right in this case:

If poss we want a generic ECN behaviour for all tunnels that will 
"just work" whether for ECN, PCN, IPsec, VPN, Mobile IP, VCP, xyz, 
whatever. Then no-one has to go round re-configuring all the tunnels 
when someone else independently deploys new feature xyz. And no-one 
has to config each new tunnel they install depending on whether there 
is also xyz in the network.

A different alarm config for decapsulators if they are within a PCN 
domain seems at odds with this goal.

But alarms are there for safety. If an operator wants safety, they 
*might* want to alarm *all* the unsafe cases. If they don't want to 
support VCP or PCN, then codepoints used by VCP or PCN are unsafe in 
their view. But this is still OK - it doesn't stop VCP or PCN being 
deployed by someone independently - they will still work, even if 
bells go off all over the place at first.

Independent deployability is the important property that we need to 
provide in a generic internetwork layer.

For instance a network operator might want to deploy VCP or PCN over 
a network they lease from a wholesaler. They don't want to have to 
depend on the wholesaler blessing what they do.


I will stick out a lot longer if anyone wants unnecessarily different 
ECN _behaviours_ in different types of tunnel.



Bob


________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From lars.eggert@nokia.com  Sun Jul 26 06:20:26 2009
Return-Path: <lars.eggert@nokia.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 6EA033A68EB for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 06:20:26 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.164
X-Spam-Level: 
X-Spam-Status: No, score=-2.164 tagged_above=-999 required=5 tests=[AWL=0.435,  BAYES_00=-2.599]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id adxOu4eTBMRm for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 06:20:25 -0700 (PDT)
Received: from mail.fit.nokia.com (mail.fit.nokia.com [195.148.124.195]) by core3.amsl.com (Postfix) with ESMTP id 4ADC43A6959 for <tsvwg@ietf.org>; Sun, 26 Jul 2009 06:20:25 -0700 (PDT)
Received: from host-217-213-50-141.homerun.telia.com (host-217-213-50-141.homerun.telia.com [217.213.50.141]) (authenticated bits=0) by mail.fit.nokia.com (8.14.3/8.14.3) with ESMTP id n6QDKDJZ047199 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=NOT); Sun, 26 Jul 2009 16:20:13 +0300 (EEST) (envelope-from lars.eggert@nokia.com)
Message-Id: <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com>
From: Lars Eggert <lars.eggert@nokia.com>
To: "James M. Polk" <jmpolk@cisco.com>
In-Reply-To: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com>
Content-Type: multipart/signed; boundary=Apple-Mail-67--275726663; micalg=sha1; protocol="application/pkcs7-signature"
Mime-Version: 1.0 (Apple Message framework v935.3)
Date: Sun, 26 Jul 2009 15:20:12 +0200
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com>
X-Mailer: Apple Mail (2.935.3)
X-Greylist: Sender succeeded SMTP AUTH, not delayed by milter-greylist-4.2.2 (mail.fit.nokia.com [212.213.221.39]); Sun, 26 Jul 2009 16:20:14 +0300 (EEST)
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 13:20:26 -0000

--Apple-Mail-67--275726663
Content-Type: text/plain;
	charset=US-ASCII;
	format=flowed;
	delsp=yes
Content-Transfer-Encoding: 7bit

Hi,

writing as an individual, with regards to the amplification attack I  
brought up as a comment from the mike at the last meeting, the draft  
now says:

>    A misbehaving Calling Node can include too many TSPECs in the
>    MULTI_TSPEC object, which can lead to an amplification attack. That
>    said, a bad implementation can create a reservation for each TSPEC
>    received from within the Resv message. The number of TSPECs in the
>    new MULTI_TSPEC object is limited, and the spec clearly states that
>    only a single reservation is to be set up per Resv message.


I think this text is inaccurate and maybe even misleading.

First, the issue isn't that a "misbehaving node" may include a few too  
many TSPECs in the packet - it's that an actively malicious attacker  
will cram as many TSPECs into an object as possible, in order to eat  
control plane cycles and data plane bandwidth.

Second, the number of TSPECs is limited in the draft, but the limit is  
17 (Section 3.2), which is a nice big amplification factor, esp.  
considering that the concern here isn't just dataplane bandwidth but  
also control plane overload. Yes, only a single reservation will be  
set up in the end, but the processing overhead goes up with the number  
of TSPECs.

Lars
--Apple-Mail-67--275726663
Content-Disposition: attachment;
	filename=smime.p7s
Content-Type: application/pkcs7-signature;
	name=smime.p7s
Content-Transfer-Encoding: base64
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==

--Apple-Mail-67--275726663--

From Black_David@emc.com  Sun Jul 26 06:28:59 2009
Return-Path: <Black_David@emc.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 624583A6887 for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 06:28:59 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -7.742
X-Spam-Level: 
X-Spam-Status: No, score=-7.742 tagged_above=-999 required=5 tests=[AWL=0.857,  BAYES_00=-2.599, GB_I_INVITATION=-2, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id PI+6W33BawSo for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 06:28:57 -0700 (PDT)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id 096A53A659C for <tsvwg@ietf.org>; Sun, 26 Jul 2009 06:28:56 -0700 (PDT)
Received: from hop04-l1d11-si03.isus.emc.com (HOP04-L1D11-SI03.isus.emc.com [10.254.111.23]) by mexforward.lss.emc.com (Switch-3.3.2/Switch-3.1.7) with ESMTP id n6QDSv77001353 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Sun, 26 Jul 2009 09:28:57 -0400
Received: from mailhub.lss.emc.com (numailhub.lss.emc.com [10.254.144.16]) by hop04-l1d11-si03.isus.emc.com (Tablus Interceptor); Sun, 26 Jul 2009 09:28:54 -0400
Received: from corpussmtp3.corp.emc.com (corpussmtp3.corp.emc.com [10.254.64.53]) by mailhub.lss.emc.com (Switch-3.3.2/Switch-3.3.2) with ESMTP id n6QDSnKD024635; Sun, 26 Jul 2009 09:28:54 -0400
Received: from CORPUSMX80A.corp.emc.com ([10.254.89.202]) by corpussmtp3.corp.emc.com with Microsoft SMTPSVC(6.0.3790.3959);  Sun, 26 Jul 2009 09:28:54 -0400
X-MimeOLE: Produced By Microsoft Exchange V6.5
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Sun, 26 Jul 2009 09:28:52 -0400
Message-ID: <9FA859626025B64FBC2AF149D97C944A034AAFDD@CORPUSMX80A.corp.emc.com>
In-Reply-To: <200907240849.n6O8nctP026517@bagheera.jungle.bt.co.uk>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
Thread-Index: AcoMO71/yLTsXKZ/QDWuao71ea0rcgBuHyvg
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk> <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AA985@CORPUSMX80A.corp.emc.com> <200907240849.n6O8nctP026517@bagheera.jungle.bt.co.uk>
From: <Black_David@emc.com>
To: <rbriscoe@jungle.bt.co.uk>
X-OriginalArrivalTime: 26 Jul 2009 13:28:54.0059 (UTC) FILETIME=[054D6FB0:01CA0DF5]
X-EMM-EM: Active
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 13:28:59 -0000

Bob,

I think you may have gone back into the PCN-like behavior swamp
by proposing to distinguish ECT(0) from ECT(1).

Alternatively, are you arguing that this distinction is
necessary in tandem with change b):
  [ECT(0) inner + ECT(1) outer -> ECT(1) in resulting header]
in order to make 3 ECN codepoints really usable?

Thanks,
--David

> -----Original Message-----
> From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]=20
> Sent: Friday, July 24, 2009 4:49 AM
> To: Black, David
> Cc: tsvwg@ietf.org
> Subject: RE: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
>=20
> David,
>=20
> Thanks for the insight into the likely reasoning.
>=20
>  From a transport perspective, I think now is the right time to say=20
> safety needs greater attention...
> - there's a desire to change 4301 decap anyway to support 2=20
> severity levels,
> - safety's become more important since 4301 - greater chance=20
> of errors...
>=20
> Guys I've talked to in the pseudowire area are considering using ECN=20
> for middlebox to middlebox congestion control (a bit like PCN).=20
> Pseudowires (or PCN) might or might not also require IPsec. But it's=20
> quite an art to turn on ECN across a tunnelled segment and do no harm=20
> in both the following cases:
> - if e2e ECN is on, avoiding polluting e2e ECN signals
> - if e2e ECN is off, avoiding turning it on
> (see recent PCN list).
>=20
> This all means the chances that an error will occur increases, either=20
> in broken implementations or broken deployments.
>=20
> The new rules in ecn-tunnel are designed so that tunnels should "just=20
> work" and "safely", even if they are deployed by someone who doesn't=20
> know they have to cater for an in-path internal ECN loop as well.
>=20
> Q. What do you think of extending drop to both CE and ECT(1) outer if=20
> there's a Not-ECT inner (see table)?
> Reasoning: As these are both currently error conditions and likely to=20
> remain so, rather than two different safety rules for different=20
> circumstances just make tunnels safe in all potential cases.=20
> Otherwise someone has to remember to configure a different safety=20
> rule on all their tunnels when PCN or pseudowires get deployed in=20
> their network.
>=20
> +----------------+----------------------------------------------+
> | Incoming Inner |             Incoming Outer Header            |
> |     Header     +---------+------------+------------+----------+
> |                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> +----------------+---------+------------+------------+----------+
> |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    | drop  =20
>   |<-3168
> |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |Not-ECT=20
>   |<-4301
> |    Not-ECT     | Not-ECT | Not-ECT    | drop       | drop  =20
>   |<-ecn-tun
> +----------------+---------+------------+------------+----------+
>                   |                Outgoing Header               |
>                   +----------------------------------------------+
>=20
>=20
> Bob
>=20
> At 22:44 23/07/2009, Black_David@emc.com wrote:
> >Bob,
> >
> > > RFC4301 seems to say: If the inner is Not-ECT, anything=20
> else in the
> > > outer must be an error. But whereas RFC3168 dropped the packet if
> > > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > > in all cases.
> > >
> > > RFC4301 gives no justification for this rather important=20
> change from
> > > RFC3168. I've spent the last hour or so trawling the=20
> IPsec and tsvwg
> > > archives and can't find any discussion on this point=20
> either. Was it
> > > discussed anywhere?
> >
> >Not that I recall.  I believe you're looking at a judgment call for
> >error case handling.  RFC 4301 was specified with close to the
> >minimum necessary to get ECN to work across an IPsec tunnel, as
> >security is an area where the default should be leaving out things
> >that aren't really necessary.  I believe that handling of this error
> >case was essentially treated as a "don't care."  Looking at your
> >alternative lines of thought:
> >
> > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > > then subsequently a congested router marked it CE, then=20
> it's unsafe
> > > to remove CE and forward Not-ECT - drop would have been safer.
> >
> >The ingress is rather unlikely to have put ECT in the outer - that
> >clearly violates both RFC 3168 and RFC 4301, and is a bad idea.
> >Hence something's almost certainly broken ...
> >
> > > * If something broken set CE on the outer even tho there was no
> > > congestion, it's safe.
> >
> >If the packet shows up with Not-ECT in the inner and CE in the
> >outer, something's definitely broken, even if the breakage is
> >actually a bad Not-ECT -> ECT transition followed by a correct
> >downstream ECT->CE transition.  Not attempting to perform a
> >compensating fix is a reasonable course of action, IMHO ...
> >but I could also defend the 3168 drop behavior as within
> >reason, along the lines that you suggest.
> >
> >Thanks,
> >--David
> >
> >
> > > -----Original Message-----
> > > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > > Sent: Monday, July 20, 2009 7:48 AM
> > > To: Black, David
> > > Cc: tsvwg@ietf.org
> > > Subject: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
> > >
> > > David,
> > >
> > > I wanted to move on from optional alarms to my proposed functional
> > > changes, ie. your case a).
> > >
> > > But I have to admit a big error which makes that moot. In
> > > ecn-tunnel-02 I incorrectly reported what RFC3168 &=20
> RFC4301 said (I
> > > just re-read them). See the table below.
> > >
> > > IP in IP Decapsulation; Recap of Pre-existing Behaviour
> > >+----------------+----------------------------------------------+
> > >| Incoming Inner |             Incoming Outer Header            |
> > >|     Header     +---------+------------+------------+----------+
> > >|                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> > >+----------------+---------+------------+------------+----------+
> > >|    Not-ECT     | Not-ECT | drop       | drop       |=20
> drop     |<my error
> > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |=20
> drop     |<-3168
> > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT   =20
> |Not-ECT   |<-4301
> > >|     ECT(0)     |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
> > >|     ECT(1)     |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
> > >|       CE       |      CE |     CE     |     CE     |   CE     |
> > >+----------------+---------+------------+------------+----------+
> > >                 |                Outgoing Header               |
> > >                 +----------------------------------------------+
> > >
> > > RFC4301 seems to say: If the inner is Not-ECT, anything=20
> else in the
> > > outer must be an error. But whereas RFC3168 dropped the packet if
> > > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > > in all cases.
> > >
> > > RFC4301 gives no justification for this rather important=20
> change from
> > > RFC3168. I've spent the last hour or so trawling the=20
> IPsec and tsvwg
> > > archives and can't find any discussion on this point=20
> either. Was it
> > > discussed anywhere?
> > >
> > > This RFC4301 behaviour seems unsafe, but perhaps=20
> understandable. It
> > > all depends how the situation arose:
> > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > > then subsequently a congested router marked it CE, then=20
> it's unsafe
> > > to remove CE and forward Not-ECT - drop would have been safer.
> > > * If something broken set CE on the outer even tho there was no
> > > congestion, it's safe.
> > >
> > >
> > >
> > > Bob
> > >
> > > At 12:07 13/07/2009, Bob Briscoe wrote:
> > > >David,
> > > >
> > > >We're getting somewhere (if we set aside (a) for now).
> > > >
> > > > > At 17:09 10/07/2009, Black_David@emc.com wrote:
> > > >> > >Quick summary of the four changes proposed in the draft:
> > > >> > >
> > > >> > >a) Inner: Not-ECT, Outer: ECT(0)
> > > >> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> > > >> > >b) Inner: ECT(0), Outer: ECT(1)
> > > >> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
> > > >> > >c) Inner: ECT(1), Outer: ECT(0)
> > > >> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > > >> > >d) Inner: CE, Outer: ECT(1)
> > > >> > >         RFC 3168/4301: CE, Change: CE(!!!)
> > > >> > >
> > > >> > >If we were designing from scratch, all of these=20
> changes would
> > > >> > >be fine things to do.  We're not, and especially after the
> > > >> > >discussion of how this impacts RFC 2003 gets added, this
> > > >> > >draft will be proposing changes to some of the most widely
> > > >> > >deployed mechanisms in the Internet.  To me, that strongly
> > > >> > >suggests a conservative design approach - focus on=20
> what really
> > > >> > >needs to be changed, and resist them temptation to fix other
> > > >> > >things that would be "nice to have".  In other words, IMHO
> > > >> > >the appropriate criteria for "fixing it" is a very strong
> > > >> > >cast that "it's broken".
> > > >> > >
> > > >> > >This is an area of principled disagreement on goals between
> > > >> > >Bob and myself that list discussion is unlikely to resolve.
> > > >> > >My view of the above four changes is that only b) makes it
> > > >> > >over the "really needs to be changed" bar (and even then,
> > > >> > >only with a *much* stronger statement of the PCN-based
> > > >> > >rationale than is in the current draft, and Bob concurs
> > > >> > >with strengthening the rationale).
> > > >> > >
> > > >> > >Turning to the other three changes, a) is clearly described
> > > >> > >as "nice to have", enough said.
> > > >> > >
> > > >> > >For c) and d), I don't believe the optional alarms are
> > > >> > >justified based on ECN (RFC 3168) alone.  A key observation
> > > >> > >is that this draft does not make the < b) ECT(0) [Inner]
> > > >> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> > > >> > >optional alarm.  What this means is that this draft not only
> > > >> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
> > > >> > >*assumes* that they are ordered in terms of priority,
> > > >> > >severity, or something analogous.  That is unwarranted as
> > > >> > >a general design assumption, IMHO.
> > > >> > >
> > > >> > >For this, reason, I think discussion of which=20
> transitions are
> > > >> > >optionally alarm-able in this fashion belongs in=20
> the context of
> > > >> > >a protocol that actually orders these codepoints, not in a
> > > >> > >general document that would serve to "just constrain future
> > > >> > >standards," borrowing Toby's words ;-).
> > > >> > >
> > > >> > >I could see discussion of optional alarms in an=20
> Appendix that
> > > >> > >discusses 3-valued PCN as an example, but not in=20
> the main body
> > > >> > >of the draft.  In the main body, it would be fine=20
> to say that
> > > >> > >specific uses of the ECN codepoints may assign meanings that
> > > >> > >cause some of these transitions to be appropriate for alarms
> > > >> > >(and point to the Appendix for a PCN example), but it would
> > > >> > >not be ok to constrain the assignment of those meanings by
> > > >> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> > > >> > >optionally alarm-able.
> > > >> > >
> > > >> > >Thanks,
> > > >> > >--David
> > > >> > >
> > > >> > >
> > > >> > > > -----Original Message-----
> > > >> > > > From: tsvwg-bounces@ietf.org=20
> [mailto:tsvwg-bounces@ietf.org]
> > > >> > > > On Behalf Of Bob Briscoe
> > > >> > > > Sent: Thursday, July 09, 2009 3:08 PM
> > > >> > > > To: Black, David
> > > >> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > > >> > > > Subject: Re: [tsvwg] review of
> > > >> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > > >> > > >
> > > >> > > > David,
> > > >> > > >
> > > >> > > > Bringing to the list as agreed. I'll reply in 2
> > > parts. Part 1/2...
> > > >> > > >
> > > >> > > > Your comments get to the nub of the proposed egress
> > > >> > protocol changes.
> > > >> > > > Great... inline...
> > > >> > > >
> > > >> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > > >> > > > >Bob,
> > > >> > > > >
> > > >> > > > >Sorry for the delay on this: things got entirely too
> > > >> > busy around the
> > > >> > > > >end of the quarter here.  Having re-read the draft from
> > > >> > scratch, I
> > > >> > > > >have a number of things to comment on ... I will be
> > > in Stockholm,
> > > >> > > > >and some of these items are going to need=20
> further discussion
> > > >> > > > >(private and/or public) in Stockholm.
> > > >> > > > >
> > > >> > > > >Starting with this table in Figure 4:
> > > >> > > > >
> > > >> > > >
> > > >> >
> > >=20
> >+------------------+----------------------------------------------+
> > > >> > > > >|  Incoming Inner  |             Incoming Outer Header
> > > >> >          |
> > > >> > > > >|      Header
> > > >> > +---------+------------+------------+----------+
> > > >> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
> > > >> >    CE    |
> > > >> > > >
> > > >> >
> > >=20
> >+------------------+---------+------------+------------+----------+
> > > >> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
> > > >> > drop(!!!)|
> > > >> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
> > > >> >   CE     |
> > > >> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
> > > >> >   CE     |
> > > >> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
> > > >> >   CE     |
> > > >> > > >
> > > >> >
> > >=20
> >+------------------+---------+------------+------------+----------+
> > > >> > > > >                    |                Outgoing Header
> > > >> >           |
> > > >> > > > >
> > > >> > +----------------------------------------------+
> > > >> > > > >
> > > >> > > > >My initial focus is on what needs to change from RFC
> > > >> > 3168/RFC 4301
> > > >> > > > >behavior and why.  Given that these two RFCs line up, I
> > > >> > think changes
> > > >> > > > >need strong motivation.  So, here are the four=20
> changes, and
> > > >> > > > my thoughts:
> > > >> > > > >
> > > >> > > > >a) Inner: Not-ECT, Outer: ECT(0)
> > > >> > > > >         3168/4301: drop (!!!)
> > > >> > > > >         Change: Not-ECT (!!!)
> > > >> > > > >         Rationale: "allow future standards actions
> > > to use this
> > > >> > > > >                 combination"
> > > >> > > > >         My view: This ain't broke, why are you=20
> fixing it?
> > > >> > > >
> > > >> > > > It is broke in the sense tunnel egresses unnecessarily
> > > >> > drop harmless
> > > >> > > > packets merely because they carry a "currently=20
> unused" (CU)
> > > >> > > > combination of codepoints. This is similar to the problem
> > > >> > ECN faced
> > > >> > > > with firewalls dropping packets unless they were Not-ECT.
> > > >> > > >
> > > >> > > > (a) wasn't originally part of the changes I was
> > > >> > proposing. But Anil
> > > >> > > > Agawaal convinced me it is worth doing. Reasoning: If we
> > > >> > are making a
> > > >> > > > change to decap anyway, let's get it as right as
> > > possible. If we
> > > >> > > > decide not to make any other changes to decap, I wouldn't
> > > >> > fight for
> > > >> > > > this one alone.
> > > >> > > >
> > > >> > > >
> > > >> > > > >b) Inner: ECT(0), Outer: ECT(1)
> > > >> > > > >         3168/4301: ECT(0)
> > > >> > > > >         Change: ECT(1)
> > > >> > > > >         Rationale: Treat ECT(1) as more severe=20
> congestion
> > > >> > > > than ECT(0).
> > > >> > > > >         My view: Longer discussion below.
> > > >> > > > >
> > > >> > > > >c) Inner: ECT(1), Outer: ECT(0)
> > > >> > > > >         3168/4301: ECT(1)
> > > >> > > > >         Change: ECT(1) (!!!)
> > > >> > > > >         Rationale: "no currently legal transition" can
> > > >> > > > produce this combo.
> > > >> > > > >         My view: Not a good idea, among other reasons
> > > >> > > > because it's at
> > > >> > > > >                 odds with your "future=20
> standards action"
> > > >> > > > explanation for
> > > >> > > > >                 the first change.  I'd remove=20
> this change
> > > >> > > > and instead deal
> > > >> > > > >                 with the possibility of=20
> raising an alarm in
> > > >> > > > a PCN-specific
> > > >> > > > >                 draft rather than declaring=20
> this to be an
> > > >> > > > alarm-able combo
> > > >> > > > >                 in general.
> > > >> > > > >
> > > >> > > > >d) Inner: CE, Outer: ECT(1)
> > > >> > > > >         3168/4301: CE
> > > >> > > > >         Change: CE (!!!)
> > > >> > > > >         Rationale & My view: <same as previous change>
> > > >> > > >
> > > >> > > > c) & d) aren't PCN specific.
> > > >> > > >
> > > >> > > > Reasoning for adding alarms (also said this eariler
> > > on list in
> > > >> > > > conversation with Gorry):
> > > >> > > > If security folks have concerns over possible attacks
> > > >> > from "currently
> > > >> > > > unused" combinations, I want to give them a way=20
> to manage the
> > > >> > > > perceived threat without just dropping packets. RFC3168
> > > >> > recommended
> > > >> > > > alarms for CU combinations. In these cases, I believe
> > > >> > RFC3168 simply
> > > >> > > > overlooked that these are also CU combinations.
> > > >> > > >
> > > >> > > >
> > > >> > > > >So, I would remove a) as not defensible (no good reason
> > > >> > for change),
> > > >> > > > >and I would push the discussion of adding alarms for c)
> > > >> > and d) into
> > > >> > > > >a PCN-specific document.
> > > >> > > >
> > > >> > > > c) & d) are not to do with PCN.
> > > >> > > >
> > > >> > > > My motivation is: While making change (b) to enable
> > > PCN, lets do
> > > >> > > > (a,c,d) to make the IP layer as clean as possible, to
> > > >> > ensure easier
> > > >> > > > future evolvability.
> > > >> > > >
> > > >> > > > I don't see the issue. It says SHOULD log and MAY
> > > alarm. So the
> > > >> > > > operator doesn't have to do anything.
> > > >> > > >
> > > >> > > > >The c) alarm in particular is only
> > > >> > > > >meaningful if PCN or something like it is in use.
> > > >> > > >
> > > >> > > > No. The combination in c) always was CU, and it's still
> > > >> > CU with PCN.
> > > >> > > > None of the proposed changes are PCN-specific.
> > > >> > > >
> > > >> > > > Even (b) is intended to also be useful for any 3-level
> > > >> > ECN while not
> > > >> > > > precluding 2-level ECN and the nonce.
> > > >> > > >
> > > >> > > > NB:  3 level =3D 2 'congested' levels + 1 level not
> > > >> > congested, e.g. VCP
> > > >> > > > or PCN which were both mentioned in RFC4774. This doesn't
> > > >> > count the
> > > >> > > > not-ECN-capable state, which makes 4 states.
> > > >> > > >
> > > >> > > > >I don't view d)
> > > >> > > > >as deserving of an alarm in general, as even though
> > > something has
> > > >> > > > >happened, no damage has been done - this is in=20
> line with the
> > > >> > > > >"be liberal in what you accept" principle of
> > > protocol design.
> > > >> > > >
> > > >> > > > That's why I'm proposing an optional alarm. Optional
> > > alarms *are*
> > > >> > > > liberal, but they allow operators a way to address
> > > >> > security concerns
> > > >> > > > about 'known unknowns' [Rumsfelt], while still=20
> allowing new
> > > >> > > > behaviours to be introduced.
> > > >> > > >
> > > >> > > >
> > > >> > > > >That leaves b), which I think is the key desired
> > > change, because
> > > >> > > > >it makes a third ECN code point usable through tunnels.
> > > >> > > >
> > > >> > > > Yes - exactly.
> > > >> > > >
> > > >> > > > >Much as
> > > >> > > > >this might be useful for PCN, it may be too late to
> > > do this.  The
> > > >> > > > >problem starts from the need to say that this draft
> > > >> > updates RFC 4301.
> > > >> > > > >The result will be a mix of 4301 and updated-4301
> > > implementations
> > > >> > > > >that render the third state difficult to use in
> > > practice.  The
> > > >> > > > >problem's even worse for non-IPsec tunnels.
> > > >> > > >
> > > >> > > > The idea is that an operator wanting to implement
> > > PCN will have a
> > > >> > > > spec they can point to in their invitation to
> > > tender. At the same
> > > >> > > > time this spec both UPDATEs and is backward compatible
> > > >> > with RFC3168 &
> > > >> > > > RFC4301 because it only changes CU stuff at egress and
> > > >> > the change at
> > > >> > > > the ingress leads to the same output at the egress.
> > > >> > Therefore vendors
> > > >> > > > can update 3168 & 4301 tunnels lazily - only when their
> > > >> > customers ask
> > > >> > > > for the new features (whether for PCN or for something
> > > >> > else like VCP).
> > > >> > > >
> > > >> > > > >I also read Appendix D
> > > >> > > > >as saying that PCN is going the two-state=20
> route, not three.
> > > >> > > >
> > > >> > > > That's not how it should be read - I'll clarify it. The
> > > >> > PCN w-g wants
> > > >> > > > 3 states, but can't have them while ecn-tunnel is
> > > not proposed
> > > >> > > > standard. Therefore, it has decided to define a 2-state
> > > >> > baseline and
> > > >> > > > a number of experimental schemes to work round this
> > > tunnel issue
> > > >> > > > (plus a comparison doc about them all). It has agreed
> > > >> > that eventually
> > > >> > > > just one experimental scheme will be chosen for
> > > standards track:
> > > >> > > > - 3-in-1 (the simplest, which depends on (b) of=20
> ecn-tunnel)
> > > >> > > > - 3-state (which uses an extra DSCP)
> > > >> > > > - psdm (which requires out of band state)
> > > >> > > > - sm (single marking gets 3 functions from 2
> > > encoding states but
> > > >> > > > admits admission control & termination aren't=20
> very precise in
> > > >> > > > all cases)
> > > >> > > >
> > > >> > > > ___
> > > >> > > > See part 2/2 for responses to your comments on=20
> specific parts
> > > >> > > > of the test.
> > > >> > > >
> > > >> > > > Thanks again for the review
> > > >> > > >
> > > >> > > >
> > > >> > > > Bob
> > > >> > > >
> > > >> > > >
> > > >> > > >
> > > >> > > >
> > > ________________________________________________________________
> > > >> > > > Bob Briscoe,               Networks Research Centre,
> > > BT Research
> > > >> > > >
> > > >> > > >
> > > >> > > >
> > > >> >
> > > >> >=20
> ________________________________________________________________
> > > >> > Bob Briscoe,               Networks Research Centre,=20
> BT Research
> > > >> >
> > > >> >
> > > >> >
> > > >
> > > >________________________________________________________________
> > > >Bob Briscoe,               Networks Research Centre, BT Research
> > >
> > > ________________________________________________________________
> > > Bob Briscoe,               Networks Research Centre, BT Research
> > >
> > >
> > >
>=20
> ________________________________________________________________
> Bob Briscoe,               Networks Research Centre, BT Research=20
>=20
>=20
>=20

From flefauch@cisco.com  Sun Jul 26 09:00:27 2009
Return-Path: <flefauch@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id C59E13A6822 for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 09:00:27 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -9.932
X-Spam-Level: 
X-Spam-Status: No, score=-9.932 tagged_above=-999 required=5 tests=[AWL=0.667,  BAYES_00=-2.599, RCVD_IN_DNSWL_HI=-8]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id g9IG8oNRiyDI for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 09:00:26 -0700 (PDT)
Received: from ams-iport-1.cisco.com (ams-iport-1.cisco.com [144.254.224.140]) by core3.amsl.com (Postfix) with ESMTP id 565753A6915 for <tsvwg@ietf.org>; Sun, 26 Jul 2009 09:00:25 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: Ak8AAF8abEqQ/uCKe2dsb2JhbACZewEBFiQGm1mIKI13BYQN
X-IronPort-AV: E=Sophos;i="4.43,272,1246838400"; d="scan'208";a="45856898"
Received: from ams-dkim-1.cisco.com ([144.254.224.138]) by ams-iport-1.cisco.com with ESMTP; 26 Jul 2009 16:00:25 +0000
Received: from ams-core-1.cisco.com (ams-core-1.cisco.com [144.254.224.150]) by ams-dkim-1.cisco.com (8.12.11/8.12.11) with ESMTP id n6QG0Pbg031248;  Sun, 26 Jul 2009 18:00:25 +0200
Received: from xbh-ams-102.cisco.com (xbh-ams-102.cisco.com [144.254.73.132]) by ams-core-1.cisco.com (8.13.8/8.14.3) with ESMTP id n6QG0P2Q015177; Sun, 26 Jul 2009 16:00:25 GMT
Received: from xfe-ams-332.cisco.com ([144.254.231.73]) by xbh-ams-102.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Sun, 26 Jul 2009 18:00:24 +0200
Received: from hakhan-unix.cisco.com ([10.61.80.157]) by xfe-ams-332.cisco.com with Microsoft SMTPSVC(6.0.3790.1830);  Sun, 26 Jul 2009 18:00:23 +0200
Message-Id: <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com>
From: Francois Le Faucheur IMAP <flefauch@cisco.com>
To: Lars Eggert <lars.eggert@nokia.com>
In-Reply-To: <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (Apple Message framework v935.3)
Date: Sun, 26 Jul 2009 18:00:11 +0200
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com>
X-Mailer: Apple Mail (2.935.3)
X-OriginalArrivalTime: 26 Jul 2009 16:00:23.0883 (UTC) FILETIME=[2F4281B0:01CA0E0A]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=1965; t=1248624025; x=1249488025; c=relaxed/simple; s=amsdkim1002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=flefauch@cisco.com; z=From:=20Francois=20Le=20Faucheur=20IMAP=20<flefauch@cisco. com> |Subject:=20Re=3A=20[tsvwg]=20New=20rev=20of=20IntServ=20MU LTI_TSPEC |Sender:=20; bh=VRe0ycx5W4oEpZzzXOwQPfYy+PlcG7tRsZGRHYknR4M=; b=dUeeg8QMx+U7dBzGmqJDGKtWmORsaGnB+4/33TE3D6ge2pFYgSF/KncEUh TNd7qGBfnDqx30ulzX9w8qc7twFIFKxTAwXVSWzgA+ZQBuvjIhdaPon1jmCi z8vajvgNMh;
Authentication-Results: ams-dkim-1; header.From=flefauch@cisco.com; dkim=pass ( sig from cisco.com/amsdkim1002 verified; ); 
Cc: "James M. Polk" <jmpolk@cisco.com>, tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 16:00:27 -0000

Hi Lars,

On 26 Jul 2009, at 15:20, Lars Eggert wrote:

> Hi,
>
> writing as an individual, with regards to the amplification attack I  
> brought up as a comment from the mike at the last meeting, the draft  
> now says:
>
>>   A misbehaving Calling Node can include too many TSPECs in the
>>   MULTI_TSPEC object, which can lead to an amplification attack. That
>>   said, a bad implementation can create a reservation for each TSPEC
>>   received from within the Resv message. The number of TSPECs in the
>>   new MULTI_TSPEC object is limited, and the spec clearly states that
>>   only a single reservation is to be set up per Resv message.
>
>
> I think this text is inaccurate and maybe even misleading.
>
> First, the issue isn't that a "misbehaving node" may include a few  
> too many TSPECs in the packet - it's that an actively malicious  
> attacker will cram as many TSPECs into an object as possible, in  
> order to eat control plane cycles and data plane bandwidth.

If you are not protected against malicious RSVP attackers, there are  
lots of very bad things you have to worry about (probably more  
worrisome than the cramming of many TSPECs in a message). For  
starters, the attacker can simply reserve all remaining bandwidth to  
all destinations. Hence the recommendation to implement neighbor  
authentication, which will protect you against malicious attackers.
So, I am not sure we need to worry too much about bad things that  
might be done maliciously specifically with the multi-Tspec.

Francois


>
> Second, the number of TSPECs is limited in the draft, but the limit  
> is 17 (Section 3.2), which is a nice big amplification factor, esp.  
> considering that the concern here isn't just dataplane bandwidth but  
> also control plane overload. Yes, only a single reservation will be  
> set up in the end, but the processing overhead goes up with the  
> number of TSPECs.
>
> Lars


From lars.eggert@nokia.com  Sun Jul 26 15:54:17 2009
Return-Path: <lars.eggert@nokia.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 696C23A684D for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 15:54:17 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.599
X-Spam-Level: 
X-Spam-Status: No, score=-2.599 tagged_above=-999 required=5 tests=[BAYES_00=-2.599]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id Kj7Kq1DY9J78 for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 15:54:16 -0700 (PDT)
Received: from mail.fit.nokia.com (mail.fit.nokia.com [195.148.124.195]) by core3.amsl.com (Postfix) with ESMTP id 5FF303A67E6 for <tsvwg@ietf.org>; Sun, 26 Jul 2009 15:54:16 -0700 (PDT)
Received: from host-78-64-88-182.homerun.telia.com (host-78-64-88-182.homerun.telia.com [78.64.88.182]) (authenticated bits=0) by mail.fit.nokia.com (8.14.3/8.14.3) with ESMTP id n6QMs40k052897 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=NOT); Mon, 27 Jul 2009 01:54:05 +0300 (EEST) (envelope-from lars.eggert@nokia.com)
Message-Id: <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com>
From: Lars Eggert <lars.eggert@nokia.com>
To: Francois Le Faucheur IMAP <flefauch@cisco.com>
In-Reply-To: <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com>
Content-Type: multipart/signed; boundary=Apple-Mail-76--241294760; micalg=sha1; protocol="application/pkcs7-signature"
Mime-Version: 1.0 (Apple Message framework v935.3)
Date: Mon, 27 Jul 2009 00:54:04 +0200
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com> <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com>
X-Mailer: Apple Mail (2.935.3)
X-Greylist: Sender succeeded SMTP AUTH, not delayed by milter-greylist-4.2.2 (mail.fit.nokia.com [212.213.221.39]); Mon, 27 Jul 2009 01:54:06 +0300 (EEST)
Cc: "James M. Polk" <jmpolk@cisco.com>, tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 22:54:17 -0000

--Apple-Mail-76--241294760
Content-Type: text/plain;
	charset=US-ASCII;
	format=flowed;
	delsp=yes
Content-Transfer-Encoding: 7bit

Hi,

On 2009-7-26, at 18:00, Francois Le Faucheur IMAP wrote:
> If you are not protected against malicious RSVP attackers, there are
> lots of very bad things you have to worry about (probably more
> worrisome than the cramming of many TSPECs in a message). For
> starters, the attacker can simply reserve all remaining bandwidth to
> all destinations. Hence the recommendation to implement neighbor
> authentication, which will protect you against malicious attackers.

OK, but then you need to require use of RFC2747, which at the moment  
isn't even mentioned in the draft.

Lars
--Apple-Mail-76--241294760
Content-Disposition: attachment;
	filename=smime.p7s
Content-Type: application/pkcs7-signature;
	name=smime.p7s
Content-Transfer-Encoding: base64

MIAGCSqGSIb3DQEHAqCAMIACAQExCzAJBgUrDgMCGgUAMIAGCSqGSIb3DQEHAQAAoIIGQDCCAvkw
ggJioAMCAQICEEi7WbMMKa2GLKFpQDaOBQEwDQYJKoZIhvcNAQEFBQAwYjELMAkGA1UEBhMCWkEx
JTAjBgNVBAoTHFRoYXd0ZSBDb25zdWx0aW5nIChQdHkpIEx0ZC4xLDAqBgNVBAMTI1RoYXd0ZSBQ
ZXJzb25hbCBGcmVlbWFpbCBJc3N1aW5nIENBMB4XDTA4MDgyMzE3NDMzOVoXDTA5MDgyMzE3NDMz
OVowXDEPMA0GA1UEBBMGRWdnZXJ0MQ0wCwYDVQQqEwRMYXJzMRQwEgYDVQQDEwtMYXJzIEVnZ2Vy
dDEkMCIGCSqGSIb3DQEJARYVbGFycy5lZ2dlcnRAbm9raWEuY29tMIIBIjANBgkqhkiG9w0BAQEF
AAOCAQ8AMIIBCgKCAQEAlwQVktJZCY89iU6jcW1XnZQN+aMgF2utCUT3H3ZKB5Jbet1SDWt0md/W
571bHjxtn9CfEJdochNL3l9f1WiJNdVbJ182557Ltx9SojqthpqtA0jKEqo2gqrf+raUj1demmo0
6ocsLqv046CrwidOp6k0RAfvkKPLhD4PD9Nk3oaZuxqBz1wY4u8Q83iWMArDeXiQxfNZnOBz5cDs
VvVjTjitm3VANkbD02tNkwl5AHw7htde4yH8hIwlfzqsAtHBEah3HyOvs9b+gHg2pFz9eS+HuotY
ZKycCweRs8NKXoCg+zAkVYi3zvZEH2VOuPlpMQMrB9+fLWg2UBsTeZ864wIDAQABozIwMDAgBgNV
HREEGTAXgRVsYXJzLmVnZ2VydEBub2tpYS5jb20wDAYDVR0TAQH/BAIwADANBgkqhkiG9w0BAQUF
AAOBgQBMdV2U+ryV5t3nuBFH19XKflodN6Bc60GBYHHY/Z0+Cl08Q75qzTt02IILBg+/YVh/fygb
6pFrOm1sFtLN7fENBfbO2VtpFjP2lGUgbXTVT5xGM6+MtqZiBI6LqexAeY6gsd/taoUfy9fZG42d
ciBA9gSGlQjjWQyG8mb5HR8L9jCCAz8wggKooAMCAQICAQ0wDQYJKoZIhvcNAQEFBQAwgdExCzAJ
BgNVBAYTAlpBMRUwEwYDVQQIEwxXZXN0ZXJuIENhcGUxEjAQBgNVBAcTCUNhcGUgVG93bjEaMBgG
A1UEChMRVGhhd3RlIENvbnN1bHRpbmcxKDAmBgNVBAsTH0NlcnRpZmljYXRpb24gU2VydmljZXMg
RGl2aXNpb24xJDAiBgNVBAMTG1RoYXd0ZSBQZXJzb25hbCBGcmVlbWFpbCBDQTErMCkGCSqGSIb3
DQEJARYccGVyc29uYWwtZnJlZW1haWxAdGhhd3RlLmNvbTAeFw0wMzA3MTcwMDAwMDBaFw0xMzA3
MTYyMzU5NTlaMGIxCzAJBgNVBAYTAlpBMSUwIwYDVQQKExxUaGF3dGUgQ29uc3VsdGluZyAoUHR5
KSBMdGQuMSwwKgYDVQQDEyNUaGF3dGUgUGVyc29uYWwgRnJlZW1haWwgSXNzdWluZyBDQTCBnzAN
BgkqhkiG9w0BAQEFAAOBjQAwgYkCgYEAxKY8VXNV+065yplaHmjAdQRwnd/p/6Me7L3N9VvyGna9
fww6YfK/Uc4B1OVQCjDXAmNaLIkVcI7dyfArhVqqP3FWy688Cwfn8R+RNiQqE88r1fOCdz0Dviv+
uxg+B79AgAJk16emu59l0cUqVIUPSAR/p7bRPGEEQB5kGXJgt/sCAwEAAaOBlDCBkTASBgNVHRMB
Af8ECDAGAQH/AgEAMEMGA1UdHwQ8MDowOKA2oDSGMmh0dHA6Ly9jcmwudGhhd3RlLmNvbS9UaGF3
dGVQZXJzb25hbEZyZWVtYWlsQ0EuY3JsMAsGA1UdDwQEAwIBBjApBgNVHREEIjAgpB4wHDEaMBgG
A1UEAxMRUHJpdmF0ZUxhYmVsMi0xMzgwDQYJKoZIhvcNAQEFBQADgYEASIzRUIPqCy7MDaNmrGcP
f6+svsIXoUOWlJ1/TCG4+DYfqi2fNi/A9BxQIJNwPP2t4WFiw9k6GX6EsZkbAMUaC4J0niVQlGLH
2ydxVyWN3amcOY6MIE9lX5Xa9/eH1sYITq726jTlEBpbNU1341YheILcIRk13iSx0x1G/11fZU8x
ggMQMIIDDAIBATB2MGIxCzAJBgNVBAYTAlpBMSUwIwYDVQQKExxUaGF3dGUgQ29uc3VsdGluZyAo
UHR5KSBMdGQuMSwwKgYDVQQDEyNUaGF3dGUgUGVyc29uYWwgRnJlZW1haWwgSXNzdWluZyBDQQIQ
SLtZswwprYYsoWlANo4FATAJBgUrDgMCGgUAoIIBbzAYBgkqhkiG9w0BCQMxCwYJKoZIhvcNAQcB
MBwGCSqGSIb3DQEJBTEPFw0wOTA3MjYyMjU0MDVaMCMGCSqGSIb3DQEJBDEWBBQup7q88eO0WAdt
BK727q5Iy/b5RzCBhQYJKwYBBAGCNxAEMXgwdjBiMQswCQYDVQQGEwJaQTElMCMGA1UEChMcVGhh
d3RlIENvbnN1bHRpbmcgKFB0eSkgTHRkLjEsMCoGA1UEAxMjVGhhd3RlIFBlcnNvbmFsIEZyZWVt
YWlsIElzc3VpbmcgQ0ECEEi7WbMMKa2GLKFpQDaOBQEwgYcGCyqGSIb3DQEJEAILMXigdjBiMQsw
CQYDVQQGEwJaQTElMCMGA1UEChMcVGhhd3RlIENvbnN1bHRpbmcgKFB0eSkgTHRkLjEsMCoGA1UE
AxMjVGhhd3RlIFBlcnNvbmFsIEZyZWVtYWlsIElzc3VpbmcgQ0ECEEi7WbMMKa2GLKFpQDaOBQEw
DQYJKoZIhvcNAQEBBQAEggEAGb0E948NYWv9BU60yK5t2HPBzQHuf2NOjgnpd1n+omrEx8Lg0lgP
o9Qbl2UfBMorcmeS7oCaX/Yu8Fawel80BVeY+rDgwZlahSgodnBj46vb32PMk9E8zCI/b2oY4q9V
gV/ZLQUp5YPMXrtM1Tej+YMSdrJicXbFQawD2k+VmKt6QxormilMmVzsHL5CYFPWNZcW6LCQ/UF0
UoFPoFS7txgHdC6ng9wZqObHjJIpFP7WIBhBhNKiEq6GWE7Q2SrE9IvwEKEPfitVf3MmEZfeF+wK
2udQ7o9O9KOe0ImDAonP0qs2y+KkK4sn0xJENUnuSIsKMC8dRS7AGwnKlQE8xgAAAAAAAA==

--Apple-Mail-76--241294760--

From jmpolk@cisco.com  Sun Jul 26 16:03:22 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 1258C3A6BE6 for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 16:03:22 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.354
X-Spam-Level: 
X-Spam-Status: No, score=-6.354 tagged_above=-999 required=5 tests=[AWL=0.245,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id A989oaxUreII for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 16:03:21 -0700 (PDT)
Received: from sj-iport-1.cisco.com (sj-iport-1.cisco.com [171.71.176.70]) by core3.amsl.com (Postfix) with ESMTP id 24DF83A6BE3 for <tsvwg@ietf.org>; Sun, 26 Jul 2009 16:03:21 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AlwGAMl9bEqrR7O6/2dsb2JhbACIR64uiCiNTwWEDQ
X-IronPort-AV: E=Sophos;i="4.43,273,1246838400"; d="scan'208";a="219303023"
Received: from sj-dkim-2.cisco.com ([171.71.179.186]) by sj-iport-1.cisco.com with ESMTP; 26 Jul 2009 23:03:22 +0000
Received: from sj-core-2.cisco.com (sj-core-2.cisco.com [171.71.177.254]) by sj-dkim-2.cisco.com (8.12.11/8.12.11) with ESMTP id n6QN3MMq014749;  Sun, 26 Jul 2009 16:03:22 -0700
Received: from xbh-sjc-231.amer.cisco.com (xbh-sjc-231.cisco.com [128.107.191.100]) by sj-core-2.cisco.com (8.13.8/8.14.3) with ESMTP id n6QN3Mo6018666; Sun, 26 Jul 2009 23:03:22 GMT
Received: from xfe-sjc-212.amer.cisco.com ([171.70.151.187]) by xbh-sjc-231.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Sun, 26 Jul 2009 16:03:22 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.9.91]) by xfe-sjc-212.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Sun, 26 Jul 2009 16:03:20 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Sun, 26 Jul 2009 18:02:58 -0500
To: Lars Eggert <lars.eggert@nokia.com>, Francois Le Faucheur IMAP <flefauch@cisco.com>
From: "James M. Polk" <jmpolk@cisco.com>
In-Reply-To: <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com>
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com> <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com> <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-212AJNACcKl00009d8b@xfe-sjc-212.amer.cisco.com>
X-OriginalArrivalTime: 26 Jul 2009 23:03:20.0927 (UTC) FILETIME=[45281EF0:01CA0E45]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=718; t=1248649402; x=1249513402; c=relaxed/simple; s=sjdkim2002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20Re=3A=20[tsvwg]=20New=20rev=20of=20IntServ=20MU LTI_TSPEC |Sender:=20; bh=gJVKg3nY/679XVNJq1PbZ2EO6gX49QCOZHPTHOwJ7iQ=; b=DfcY0LhRVLeDJdxviQ4Uubnfo68PFyI8hT06tfPIpL3Sa+bIU/EBu+x/Qe pyWk9k2rNgDu7DllERFEe/BO4NyKbiY842PDMhgCNM4nN/kccXygHEYHlOSh +m3IU90JNb;
Authentication-Results: sj-dkim-2; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim2002 verified; ); 
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 23:03:22 -0000

At 05:54 PM 7/26/2009, Lars Eggert wrote:
>Hi,
>
>On 2009-7-26, at 18:00, Francois Le Faucheur IMAP wrote:
>>If you are not protected against malicious RSVP attackers, there are
>>lots of very bad things you have to worry about (probably more
>>worrisome than the cramming of many TSPECs in a message). For
>>starters, the attacker can simply reserve all remaining bandwidth to
>>all destinations. Hence the recommendation to implement neighbor
>>authentication, which will protect you against malicious attackers.
>
>OK, but then you need to require use of RFC2747, which at the moment
>isn't even mentioned in the draft.

sorry for missing this.

I'll get it in the next rev.

James


>Lars
>


From rbriscoe@jungle.bt.co.uk  Sun Jul 26 16:29:08 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 802313A67A2 for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 16:29:08 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.905
X-Spam-Level: 
X-Spam-Status: No, score=-2.905 tagged_above=-999 required=5 tests=[AWL=1.212,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id zSlvZHKGVp2W for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 16:29:02 -0700 (PDT)
Received: from smtp3.smtp.bt.com (smtp3.smtp.bt.com [217.32.164.138]) by core3.amsl.com (Postfix) with ESMTP id 4BECE3A6765 for <tsvwg@ietf.org>; Sun, 26 Jul 2009 16:28:38 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp3.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 00:28:38 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 27 Jul 2009 00:28:38 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248650917313; Mon, 27 Jul 2009 00:28:37 +0100
Received: from mut.jungle.bt.co.uk ([10.73.144.52]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6QNSVD9000689; Mon, 27 Jul 2009 00:28:33 +0100
Message-Id: <200907262328.n6QNSVD9000689@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 27 Jul 2009 00:28:13 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A034AAFDD@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk> <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AA985@CORPUSMX80A.corp.emc.com> <200907240849.n6O8nctP026517@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AAFDD@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 26 Jul 2009 23:28:38.0526 (UTC) FILETIME=[CDB759E0:01CA0E48]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 23:29:08 -0000

David,

The whole point of doing this is because we want to support PCN-like 
behaviour in combinations of codepoints that weren't otherwise used. 
So we are proposing to change the ECN behaviour for combinations that 
aren't otherwise used, so PCN can use them.

We've worked out a single ECN tunnel behaviour that's good for IPsec, 
mobile IP, VPNs, PCN, any 2-severity level ECN scheme, any 1-severity 
level ECN scheme, with a nonce or without a nonce.

This is an update to RFC3168 & RFC4301 tunnels that a network 
operator can require compliance with if it wants to support 
2-severity level ECN (like PCN).

It's backwards compatible, so any other tunnel can be updated to this 
(or not), and the previous ECN behaviour will still work as before.

Tunnels get deployed in domains independently of congestion signalling schemes.
So we don't want different tunnels for different purposes.
One tunnel behaviour for all purposes. Please.



Bob


At 14:28 26/07/2009, Black_David@emc.com wrote:
>Bob,
>
>I think you may have gone back into the PCN-like behavior swamp
>by proposing to distinguish ECT(0) from ECT(1).
>
>Alternatively, are you arguing that this distinction is
>necessary in tandem with change b):
>   [ECT(0) inner + ECT(1) outer -> ECT(1) in resulting header]
>in order to make 3 ECN codepoints really usable?
>
>Thanks,
>--David
>
> > -----Original Message-----
> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > Sent: Friday, July 24, 2009 4:49 AM
> > To: Black, David
> > Cc: tsvwg@ietf.org
> > Subject: RE: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
> >
> > David,
> >
> > Thanks for the insight into the likely reasoning.
> >
> >  From a transport perspective, I think now is the right time to say
> > safety needs greater attention...
> > - there's a desire to change 4301 decap anyway to support 2
> > severity levels,
> > - safety's become more important since 4301 - greater chance
> > of errors...
> >
> > Guys I've talked to in the pseudowire area are considering using ECN
> > for middlebox to middlebox congestion control (a bit like PCN).
> > Pseudowires (or PCN) might or might not also require IPsec. But it's
> > quite an art to turn on ECN across a tunnelled segment and do no harm
> > in both the following cases:
> > - if e2e ECN is on, avoiding polluting e2e ECN signals
> > - if e2e ECN is off, avoiding turning it on
> > (see recent PCN list).
> >
> > This all means the chances that an error will occur increases, either
> > in broken implementations or broken deployments.
> >
> > The new rules in ecn-tunnel are designed so that tunnels should "just
> > work" and "safely", even if they are deployed by someone who doesn't
> > know they have to cater for an in-path internal ECN loop as well.
> >
> > Q. What do you think of extending drop to both CE and ECT(1) outer if
> > there's a Not-ECT inner (see table)?
> > Reasoning: As these are both currently error conditions and likely to
> > remain so, rather than two different safety rules for different
> > circumstances just make tunnels safe in all potential cases.
> > Otherwise someone has to remember to configure a different safety
> > rule on all their tunnels when PCN or pseudowires get deployed in
> > their network.
> >
> > +----------------+----------------------------------------------+
> > | Incoming Inner |             Incoming Outer Header            |
> > |     Header     +---------+------------+------------+----------+
> > |                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> > +----------------+---------+------------+------------+----------+
> > |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    | drop
> >   |<-3168
> > |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |Not-ECT
> >   |<-4301
> > |    Not-ECT     | Not-ECT | Not-ECT    | drop       | drop
> >   |<-ecn-tun
> > +----------------+---------+------------+------------+----------+
> >                   |                Outgoing Header               |
> >                   +----------------------------------------------+
> >
> >
> > Bob
> >
> > At 22:44 23/07/2009, Black_David@emc.com wrote:
> > >Bob,
> > >
> > > > RFC4301 seems to say: If the inner is Not-ECT, anything
> > else in the
> > > > outer must be an error. But whereas RFC3168 dropped the packet if
> > > > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > > > in all cases.
> > > >
> > > > RFC4301 gives no justification for this rather important
> > change from
> > > > RFC3168. I've spent the last hour or so trawling the
> > IPsec and tsvwg
> > > > archives and can't find any discussion on this point
> > either. Was it
> > > > discussed anywhere?
> > >
> > >Not that I recall.  I believe you're looking at a judgment call for
> > >error case handling.  RFC 4301 was specified with close to the
> > >minimum necessary to get ECN to work across an IPsec tunnel, as
> > >security is an area where the default should be leaving out things
> > >that aren't really necessary.  I believe that handling of this error
> > >case was essentially treated as a "don't care."  Looking at your
> > >alternative lines of thought:
> > >
> > > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > > > then subsequently a congested router marked it CE, then
> > it's unsafe
> > > > to remove CE and forward Not-ECT - drop would have been safer.
> > >
> > >The ingress is rather unlikely to have put ECT in the outer - that
> > >clearly violates both RFC 3168 and RFC 4301, and is a bad idea.
> > >Hence something's almost certainly broken ...
> > >
> > > > * If something broken set CE on the outer even tho there was no
> > > > congestion, it's safe.
> > >
> > >If the packet shows up with Not-ECT in the inner and CE in the
> > >outer, something's definitely broken, even if the breakage is
> > >actually a bad Not-ECT -> ECT transition followed by a correct
> > >downstream ECT->CE transition.  Not attempting to perform a
> > >compensating fix is a reasonable course of action, IMHO ...
> > >but I could also defend the 3168 drop behavior as within
> > >reason, along the lines that you suggest.
> > >
> > >Thanks,
> > >--David
> > >
> > >
> > > > -----Original Message-----
> > > > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > > > Sent: Monday, July 20, 2009 7:48 AM
> > > > To: Black, David
> > > > Cc: tsvwg@ietf.org
> > > > Subject: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
> > > >
> > > > David,
> > > >
> > > > I wanted to move on from optional alarms to my proposed functional
> > > > changes, ie. your case a).
> > > >
> > > > But I have to admit a big error which makes that moot. In
> > > > ecn-tunnel-02 I incorrectly reported what RFC3168 &
> > RFC4301 said (I
> > > > just re-read them). See the table below.
> > > >
> > > > IP in IP Decapsulation; Recap of Pre-existing Behaviour
> > > >+----------------+----------------------------------------------+
> > > >| Incoming Inner |             Incoming Outer Header            |
> > > >|     Header     +---------+------------+------------+----------+
> > > >|                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> > > >+----------------+---------+------------+------------+----------+
> > > >|    Not-ECT     | Not-ECT | drop       | drop       |
> > drop     |<my error
> > > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |
> > drop     |<-3168
> > > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT
> > |Not-ECT   |<-4301
> > > >|     ECT(0)     |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
> > > >|     ECT(1)     |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
> > > >|       CE       |      CE |     CE     |     CE     |   CE     |
> > > >+----------------+---------+------------+------------+----------+
> > > >                 |                Outgoing Header               |
> > > >                 +----------------------------------------------+
> > > >
> > > > RFC4301 seems to say: If the inner is Not-ECT, anything
> > else in the
> > > > outer must be an error. But whereas RFC3168 dropped the packet if
> > > > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > > > in all cases.
> > > >
> > > > RFC4301 gives no justification for this rather important
> > change from
> > > > RFC3168. I've spent the last hour or so trawling the
> > IPsec and tsvwg
> > > > archives and can't find any discussion on this point
> > either. Was it
> > > > discussed anywhere?
> > > >
> > > > This RFC4301 behaviour seems unsafe, but perhaps
> > understandable. It
> > > > all depends how the situation arose:
> > > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > > > then subsequently a congested router marked it CE, then
> > it's unsafe
> > > > to remove CE and forward Not-ECT - drop would have been safer.
> > > > * If something broken set CE on the outer even tho there was no
> > > > congestion, it's safe.
> > > >
> > > >
> > > >
> > > > Bob
> > > >
> > > > At 12:07 13/07/2009, Bob Briscoe wrote:
> > > > >David,
> > > > >
> > > > >We're getting somewhere (if we set aside (a) for now).
> > > > >
> > > > > > At 17:09 10/07/2009, Black_David@emc.com wrote:
> > > > >> > >Quick summary of the four changes proposed in the draft:
> > > > >> > >
> > > > >> > >a) Inner: Not-ECT, Outer: ECT(0)
> > > > >> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> > > > >> > >b) Inner: ECT(0), Outer: ECT(1)
> > > > >> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
> > > > >> > >c) Inner: ECT(1), Outer: ECT(0)
> > > > >> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > > > >> > >d) Inner: CE, Outer: ECT(1)
> > > > >> > >         RFC 3168/4301: CE, Change: CE(!!!)
> > > > >> > >
> > > > >> > >If we were designing from scratch, all of these
> > changes would
> > > > >> > >be fine things to do.  We're not, and especially after the
> > > > >> > >discussion of how this impacts RFC 2003 gets added, this
> > > > >> > >draft will be proposing changes to some of the most widely
> > > > >> > >deployed mechanisms in the Internet.  To me, that strongly
> > > > >> > >suggests a conservative design approach - focus on
> > what really
> > > > >> > >needs to be changed, and resist them temptation to fix other
> > > > >> > >things that would be "nice to have".  In other words, IMHO
> > > > >> > >the appropriate criteria for "fixing it" is a very strong
> > > > >> > >cast that "it's broken".
> > > > >> > >
> > > > >> > >This is an area of principled disagreement on goals between
> > > > >> > >Bob and myself that list discussion is unlikely to resolve.
> > > > >> > >My view of the above four changes is that only b) makes it
> > > > >> > >over the "really needs to be changed" bar (and even then,
> > > > >> > >only with a *much* stronger statement of the PCN-based
> > > > >> > >rationale than is in the current draft, and Bob concurs
> > > > >> > >with strengthening the rationale).
> > > > >> > >
> > > > >> > >Turning to the other three changes, a) is clearly described
> > > > >> > >as "nice to have", enough said.
> > > > >> > >
> > > > >> > >For c) and d), I don't believe the optional alarms are
> > > > >> > >justified based on ECN (RFC 3168) alone.  A key observation
> > > > >> > >is that this draft does not make the < b) ECT(0) [Inner]
> > > > >> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> > > > >> > >optional alarm.  What this means is that this draft not only
> > > > >> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
> > > > >> > >*assumes* that they are ordered in terms of priority,
> > > > >> > >severity, or something analogous.  That is unwarranted as
> > > > >> > >a general design assumption, IMHO.
> > > > >> > >
> > > > >> > >For this, reason, I think discussion of which
> > transitions are
> > > > >> > >optionally alarm-able in this fashion belongs in
> > the context of
> > > > >> > >a protocol that actually orders these codepoints, not in a
> > > > >> > >general document that would serve to "just constrain future
> > > > >> > >standards," borrowing Toby's words ;-).
> > > > >> > >
> > > > >> > >I could see discussion of optional alarms in an
> > Appendix that
> > > > >> > >discusses 3-valued PCN as an example, but not in
> > the main body
> > > > >> > >of the draft.  In the main body, it would be fine
> > to say that
> > > > >> > >specific uses of the ECN codepoints may assign meanings that
> > > > >> > >cause some of these transitions to be appropriate for alarms
> > > > >> > >(and point to the Appendix for a PCN example), but it would
> > > > >> > >not be ok to constrain the assignment of those meanings by
> > > > >> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> > > > >> > >optionally alarm-able.
> > > > >> > >
> > > > >> > >Thanks,
> > > > >> > >--David
> > > > >> > >
> > > > >> > >
> > > > >> > > > -----Original Message-----
> > > > >> > > > From: tsvwg-bounces@ietf.org
> > [mailto:tsvwg-bounces@ietf.org]
> > > > >> > > > On Behalf Of Bob Briscoe
> > > > >> > > > Sent: Thursday, July 09, 2009 3:08 PM
> > > > >> > > > To: Black, David
> > > > >> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > > > >> > > > Subject: Re: [tsvwg] review of
> > > > >> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > > > >> > > >
> > > > >> > > > David,
> > > > >> > > >
> > > > >> > > > Bringing to the list as agreed. I'll reply in 2
> > > > parts. Part 1/2...
> > > > >> > > >
> > > > >> > > > Your comments get to the nub of the proposed egress
> > > > >> > protocol changes.
> > > > >> > > > Great... inline...
> > > > >> > > >
> > > > >> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > > > >> > > > >Bob,
> > > > >> > > > >
> > > > >> > > > >Sorry for the delay on this: things got entirely too
> > > > >> > busy around the
> > > > >> > > > >end of the quarter here.  Having re-read the draft from
> > > > >> > scratch, I
> > > > >> > > > >have a number of things to comment on ... I will be
> > > > in Stockholm,
> > > > >> > > > >and some of these items are going to need
> > further discussion
> > > > >> > > > >(private and/or public) in Stockholm.
> > > > >> > > > >
> > > > >> > > > >Starting with this table in Figure 4:
> > > > >> > > > >
> > > > >> > > >
> > > > >> >
> > > >
> > >+------------------+----------------------------------------------+
> > > > >> > > > >|  Incoming Inner  |             Incoming Outer Header
> > > > >> >          |
> > > > >> > > > >|      Header
> > > > >> > +---------+------------+------------+----------+
> > > > >> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
> > > > >> >    CE    |
> > > > >> > > >
> > > > >> >
> > > >
> > >+------------------+---------+------------+------------+----------+
> > > > >> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
> > > > >> > drop(!!!)|
> > > > >> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
> > > > >> >   CE     |
> > > > >> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
> > > > >> >   CE     |
> > > > >> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
> > > > >> >   CE     |
> > > > >> > > >
> > > > >> >
> > > >
> > >+------------------+---------+------------+------------+----------+
> > > > >> > > > >                    |                Outgoing Header
> > > > >> >           |
> > > > >> > > > >
> > > > >> > +----------------------------------------------+
> > > > >> > > > >
> > > > >> > > > >My initial focus is on what needs to change from RFC
> > > > >> > 3168/RFC 4301
> > > > >> > > > >behavior and why.  Given that these two RFCs line up, I
> > > > >> > think changes
> > > > >> > > > >need strong motivation.  So, here are the four
> > changes, and
> > > > >> > > > my thoughts:
> > > > >> > > > >
> > > > >> > > > >a) Inner: Not-ECT, Outer: ECT(0)
> > > > >> > > > >         3168/4301: drop (!!!)
> > > > >> > > > >         Change: Not-ECT (!!!)
> > > > >> > > > >         Rationale: "allow future standards actions
> > > > to use this
> > > > >> > > > >                 combination"
> > > > >> > > > >         My view: This ain't broke, why are you
> > fixing it?
> > > > >> > > >
> > > > >> > > > It is broke in the sense tunnel egresses unnecessarily
> > > > >> > drop harmless
> > > > >> > > > packets merely because they carry a "currently
> > unused" (CU)
> > > > >> > > > combination of codepoints. This is similar to the problem
> > > > >> > ECN faced
> > > > >> > > > with firewalls dropping packets unless they were Not-ECT.
> > > > >> > > >
> > > > >> > > > (a) wasn't originally part of the changes I was
> > > > >> > proposing. But Anil
> > > > >> > > > Agawaal convinced me it is worth doing. Reasoning: If we
> > > > >> > are making a
> > > > >> > > > change to decap anyway, let's get it as right as
> > > > possible. If we
> > > > >> > > > decide not to make any other changes to decap, I wouldn't
> > > > >> > fight for
> > > > >> > > > this one alone.
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > >b) Inner: ECT(0), Outer: ECT(1)
> > > > >> > > > >         3168/4301: ECT(0)
> > > > >> > > > >         Change: ECT(1)
> > > > >> > > > >         Rationale: Treat ECT(1) as more severe
> > congestion
> > > > >> > > > than ECT(0).
> > > > >> > > > >         My view: Longer discussion below.
> > > > >> > > > >
> > > > >> > > > >c) Inner: ECT(1), Outer: ECT(0)
> > > > >> > > > >         3168/4301: ECT(1)
> > > > >> > > > >         Change: ECT(1) (!!!)
> > > > >> > > > >         Rationale: "no currently legal transition" can
> > > > >> > > > produce this combo.
> > > > >> > > > >         My view: Not a good idea, among other reasons
> > > > >> > > > because it's at
> > > > >> > > > >                 odds with your "future
> > standards action"
> > > > >> > > > explanation for
> > > > >> > > > >                 the first change.  I'd remove
> > this change
> > > > >> > > > and instead deal
> > > > >> > > > >                 with the possibility of
> > raising an alarm in
> > > > >> > > > a PCN-specific
> > > > >> > > > >                 draft rather than declaring
> > this to be an
> > > > >> > > > alarm-able combo
> > > > >> > > > >                 in general.
> > > > >> > > > >
> > > > >> > > > >d) Inner: CE, Outer: ECT(1)
> > > > >> > > > >         3168/4301: CE
> > > > >> > > > >         Change: CE (!!!)
> > > > >> > > > >         Rationale & My view: <same as previous change>
> > > > >> > > >
> > > > >> > > > c) & d) aren't PCN specific.
> > > > >> > > >
> > > > >> > > > Reasoning for adding alarms (also said this eariler
> > > > on list in
> > > > >> > > > conversation with Gorry):
> > > > >> > > > If security folks have concerns over possible attacks
> > > > >> > from "currently
> > > > >> > > > unused" combinations, I want to give them a way
> > to manage the
> > > > >> > > > perceived threat without just dropping packets. RFC3168
> > > > >> > recommended
> > > > >> > > > alarms for CU combinations. In these cases, I believe
> > > > >> > RFC3168 simply
> > > > >> > > > overlooked that these are also CU combinations.
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > >So, I would remove a) as not defensible (no good reason
> > > > >> > for change),
> > > > >> > > > >and I would push the discussion of adding alarms for c)
> > > > >> > and d) into
> > > > >> > > > >a PCN-specific document.
> > > > >> > > >
> > > > >> > > > c) & d) are not to do with PCN.
> > > > >> > > >
> > > > >> > > > My motivation is: While making change (b) to enable
> > > > PCN, lets do
> > > > >> > > > (a,c,d) to make the IP layer as clean as possible, to
> > > > >> > ensure easier
> > > > >> > > > future evolvability.
> > > > >> > > >
> > > > >> > > > I don't see the issue. It says SHOULD log and MAY
> > > > alarm. So the
> > > > >> > > > operator doesn't have to do anything.
> > > > >> > > >
> > > > >> > > > >The c) alarm in particular is only
> > > > >> > > > >meaningful if PCN or something like it is in use.
> > > > >> > > >
> > > > >> > > > No. The combination in c) always was CU, and it's still
> > > > >> > CU with PCN.
> > > > >> > > > None of the proposed changes are PCN-specific.
> > > > >> > > >
> > > > >> > > > Even (b) is intended to also be useful for any 3-level
> > > > >> > ECN while not
> > > > >> > > > precluding 2-level ECN and the nonce.
> > > > >> > > >
> > > > >> > > > NB:  3 level = 2 'congested' levels + 1 level not
> > > > >> > congested, e.g. VCP
> > > > >> > > > or PCN which were both mentioned in RFC4774. This doesn't
> > > > >> > count the
> > > > >> > > > not-ECN-capable state, which makes 4 states.
> > > > >> > > >
> > > > >> > > > >I don't view d)
> > > > >> > > > >as deserving of an alarm in general, as even though
> > > > something has
> > > > >> > > > >happened, no damage has been done - this is in
> > line with the
> > > > >> > > > >"be liberal in what you accept" principle of
> > > > protocol design.
> > > > >> > > >
> > > > >> > > > That's why I'm proposing an optional alarm. Optional
> > > > alarms *are*
> > > > >> > > > liberal, but they allow operators a way to address
> > > > >> > security concerns
> > > > >> > > > about 'known unknowns' [Rumsfelt], while still
> > allowing new
> > > > >> > > > behaviours to be introduced.
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > >That leaves b), which I think is the key desired
> > > > change, because
> > > > >> > > > >it makes a third ECN code point usable through tunnels.
> > > > >> > > >
> > > > >> > > > Yes - exactly.
> > > > >> > > >
> > > > >> > > > >Much as
> > > > >> > > > >this might be useful for PCN, it may be too late to
> > > > do this.  The
> > > > >> > > > >problem starts from the need to say that this draft
> > > > >> > updates RFC 4301.
> > > > >> > > > >The result will be a mix of 4301 and updated-4301
> > > > implementations
> > > > >> > > > >that render the third state difficult to use in
> > > > practice.  The
> > > > >> > > > >problem's even worse for non-IPsec tunnels.
> > > > >> > > >
> > > > >> > > > The idea is that an operator wanting to implement
> > > > PCN will have a
> > > > >> > > > spec they can point to in their invitation to
> > > > tender. At the same
> > > > >> > > > time this spec both UPDATEs and is backward compatible
> > > > >> > with RFC3168 &
> > > > >> > > > RFC4301 because it only changes CU stuff at egress and
> > > > >> > the change at
> > > > >> > > > the ingress leads to the same output at the egress.
> > > > >> > Therefore vendors
> > > > >> > > > can update 3168 & 4301 tunnels lazily - only when their
> > > > >> > customers ask
> > > > >> > > > for the new features (whether for PCN or for something
> > > > >> > else like VCP).
> > > > >> > > >
> > > > >> > > > >I also read Appendix D
> > > > >> > > > >as saying that PCN is going the two-state
> > route, not three.
> > > > >> > > >
> > > > >> > > > That's not how it should be read - I'll clarify it. The
> > > > >> > PCN w-g wants
> > > > >> > > > 3 states, but can't have them while ecn-tunnel is
> > > > not proposed
> > > > >> > > > standard. Therefore, it has decided to define a 2-state
> > > > >> > baseline and
> > > > >> > > > a number of experimental schemes to work round this
> > > > tunnel issue
> > > > >> > > > (plus a comparison doc about them all). It has agreed
> > > > >> > that eventually
> > > > >> > > > just one experimental scheme will be chosen for
> > > > standards track:
> > > > >> > > > - 3-in-1 (the simplest, which depends on (b) of
> > ecn-tunnel)
> > > > >> > > > - 3-state (which uses an extra DSCP)
> > > > >> > > > - psdm (which requires out of band state)
> > > > >> > > > - sm (single marking gets 3 functions from 2
> > > > encoding states but
> > > > >> > > > admits admission control & termination aren't
> > very precise in
> > > > >> > > > all cases)
> > > > >> > > >
> > > > >> > > > ___
> > > > >> > > > See part 2/2 for responses to your comments on
> > specific parts
> > > > >> > > > of the test.
> > > > >> > > >
> > > > >> > > > Thanks again for the review
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > Bob
> > > > >> > > >
> > > > >> > > >
> > > > >> > > >
> > > > >> > > >
> > > > ________________________________________________________________
> > > > >> > > > Bob Briscoe,               Networks Research Centre,
> > > > BT Research
> > > > >> > > >
> > > > >> > > >
> > > > >> > > >
> > > > >> >
> > > > >> >
> > ________________________________________________________________
> > > > >> > Bob Briscoe,               Networks Research Centre,
> > BT Research
> > > > >> >
> > > > >> >
> > > > >> >
> > > > >
> > > > >________________________________________________________________
> > > > >Bob Briscoe,               Networks Research Centre, BT Research
> > > >
> > > > ________________________________________________________________
> > > > Bob Briscoe,               Networks Research Centre, BT Research
> > > >
> > > >
> > > >
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From rbriscoe@jungle.bt.co.uk  Sun Jul 26 16:36:16 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 1FE363A68CF for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 16:36:16 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.976
X-Spam-Level: 
X-Spam-Status: No, score=-2.976 tagged_above=-999 required=5 tests=[AWL=1.141,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 9dHSei0hIr72 for <tsvwg@core3.amsl.com>; Sun, 26 Jul 2009 16:36:09 -0700 (PDT)
Received: from smtp1.smtp.bt.com (smtp1.smtp.bt.com [217.32.164.137]) by core3.amsl.com (Postfix) with ESMTP id 5F8F73A6849 for <tsvwg@ietf.org>; Sun, 26 Jul 2009 16:36:08 -0700 (PDT)
Received: from i2kc06-ukbr.domain1.systemhost.net ([193.113.197.70]) by smtp1.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 00:36:08 +0100
Received: from cbibipnt05.iuser.iroot.adidom.com ([147.149.196.177]) by i2kc06-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Mon, 27 Jul 2009 00:36:08 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt05.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248651367514; Mon, 27 Jul 2009 00:36:07 +0100
Received: from mut.jungle.bt.co.uk ([10.73.144.52]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6QNa1AT000883; Mon, 27 Jul 2009 00:36:02 +0100
Message-Id: <200907262336.n6QNa1AT000883@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 27 Jul 2009 00:35:59 +0100
To: <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
In-Reply-To: <9FA859626025B64FBC2AF149D97C944A034AAFDD@CORPUSMX80A.corp. emc.com>
References: <200907091908.n69J838L011064@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E672F@CORPUSMX80A.corp.emc.com> <200907101731.n6AHVfRf015097@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A032E6938@CORPUSMX80A.corp.emc.com> <7.1.0.9.0.20090713110743.07650538@jungle.bt.co.uk> <200907201148.n6KBmHDt009985@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AA985@CORPUSMX80A.corp.emc.com> <200907240849.n6O8nctP026517@bagheera.jungle.bt.co.uk> <9FA859626025B64FBC2AF149D97C944A034AAFDD@CORPUSMX80A.corp.emc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 26 Jul 2009 23:36:08.0487 (UTC) FILETIME=[D9E9F370:01CA0E49]
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Sun, 26 Jul 2009 23:36:16 -0000

David,

offlist

At 14:28 26/07/2009, Black_David@emc.com wrote:
>Bob,
>
>I think you may have gone back into the PCN-like behavior swamp
>by proposing to distinguish ECT(0) from ECT(1).
>
>Alternatively, are you arguing that this distinction is
>necessary in tandem with change b):
>   [ECT(0) inner + ECT(1) outer -> ECT(1) in resulting header]
>in order to make 3 ECN codepoints really usable?

Of course. All the MUSTs in any draft should be read as a set of 
mutually supporting changes.




Bob


>Thanks,
>--David
>
> > -----Original Message-----
> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > Sent: Friday, July 24, 2009 4:49 AM
> > To: Black, David
> > Cc: tsvwg@ietf.org
> > Subject: RE: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
> >
> > David,
> >
> > Thanks for the insight into the likely reasoning.
> >
> >  From a transport perspective, I think now is the right time to say
> > safety needs greater attention...
> > - there's a desire to change 4301 decap anyway to support 2
> > severity levels,
> > - safety's become more important since 4301 - greater chance
> > of errors...
> >
> > Guys I've talked to in the pseudowire area are considering using ECN
> > for middlebox to middlebox congestion control (a bit like PCN).
> > Pseudowires (or PCN) might or might not also require IPsec. But it's
> > quite an art to turn on ECN across a tunnelled segment and do no harm
> > in both the following cases:
> > - if e2e ECN is on, avoiding polluting e2e ECN signals
> > - if e2e ECN is off, avoiding turning it on
> > (see recent PCN list).
> >
> > This all means the chances that an error will occur increases, either
> > in broken implementations or broken deployments.
> >
> > The new rules in ecn-tunnel are designed so that tunnels should "just
> > work" and "safely", even if they are deployed by someone who doesn't
> > know they have to cater for an in-path internal ECN loop as well.
> >
> > Q. What do you think of extending drop to both CE and ECT(1) outer if
> > there's a Not-ECT inner (see table)?
> > Reasoning: As these are both currently error conditions and likely to
> > remain so, rather than two different safety rules for different
> > circumstances just make tunnels safe in all potential cases.
> > Otherwise someone has to remember to configure a different safety
> > rule on all their tunnels when PCN or pseudowires get deployed in
> > their network.
> >
> > +----------------+----------------------------------------------+
> > | Incoming Inner |             Incoming Outer Header            |
> > |     Header     +---------+------------+------------+----------+
> > |                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> > +----------------+---------+------------+------------+----------+
> > |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    | drop
> >   |<-3168
> > |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |Not-ECT
> >   |<-4301
> > |    Not-ECT     | Not-ECT | Not-ECT    | drop       | drop
> >   |<-ecn-tun
> > +----------------+---------+------------+------------+----------+
> >                   |                Outgoing Header               |
> >                   +----------------------------------------------+
> >
> >
> > Bob
> >
> > At 22:44 23/07/2009, Black_David@emc.com wrote:
> > >Bob,
> > >
> > > > RFC4301 seems to say: If the inner is Not-ECT, anything
> > else in the
> > > > outer must be an error. But whereas RFC3168 dropped the packet if
> > > > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > > > in all cases.
> > > >
> > > > RFC4301 gives no justification for this rather important
> > change from
> > > > RFC3168. I've spent the last hour or so trawling the
> > IPsec and tsvwg
> > > > archives and can't find any discussion on this point
> > either. Was it
> > > > discussed anywhere?
> > >
> > >Not that I recall.  I believe you're looking at a judgment call for
> > >error case handling.  RFC 4301 was specified with close to the
> > >minimum necessary to get ECN to work across an IPsec tunnel, as
> > >security is an area where the default should be leaving out things
> > >that aren't really necessary.  I believe that handling of this error
> > >case was essentially treated as a "don't care."  Looking at your
> > >alternative lines of thought:
> > >
> > > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > > > then subsequently a congested router marked it CE, then
> > it's unsafe
> > > > to remove CE and forward Not-ECT - drop would have been safer.
> > >
> > >The ingress is rather unlikely to have put ECT in the outer - that
> > >clearly violates both RFC 3168 and RFC 4301, and is a bad idea.
> > >Hence something's almost certainly broken ...
> > >
> > > > * If something broken set CE on the outer even tho there was no
> > > > congestion, it's safe.
> > >
> > >If the packet shows up with Not-ECT in the inner and CE in the
> > >outer, something's definitely broken, even if the breakage is
> > >actually a bad Not-ECT -> ECT transition followed by a correct
> > >downstream ECT->CE transition.  Not attempting to perform a
> > >compensating fix is a reasonable course of action, IMHO ...
> > >but I could also defend the 3168 drop behavior as within
> > >reason, along the lines that you suggest.
> > >
> > >Thanks,
> > >--David
> > >
> > >
> > > > -----Original Message-----
> > > > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
> > > > Sent: Monday, July 20, 2009 7:48 AM
> > > > To: Black, David
> > > > Cc: tsvwg@ietf.org
> > > > Subject: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
> > > >
> > > > David,
> > > >
> > > > I wanted to move on from optional alarms to my proposed functional
> > > > changes, ie. your case a).
> > > >
> > > > But I have to admit a big error which makes that moot. In
> > > > ecn-tunnel-02 I incorrectly reported what RFC3168 &
> > RFC4301 said (I
> > > > just re-read them). See the table below.
> > > >
> > > > IP in IP Decapsulation; Recap of Pre-existing Behaviour
> > > >+----------------+----------------------------------------------+
> > > >| Incoming Inner |             Incoming Outer Header            |
> > > >|     Header     +---------+------------+------------+----------+
> > > >|                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
> > > >+----------------+---------+------------+------------+----------+
> > > >|    Not-ECT     | Not-ECT | drop       | drop       |
> > drop     |<my error
> > > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |
> > drop     |<-3168
> > > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT
> > |Not-ECT   |<-4301
> > > >|     ECT(0)     |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
> > > >|     ECT(1)     |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
> > > >|       CE       |      CE |     CE     |     CE     |   CE     |
> > > >+----------------+---------+------------+------------+----------+
> > > >                 |                Outgoing Header               |
> > > >                 +----------------------------------------------+
> > > >
> > > > RFC4301 seems to say: If the inner is Not-ECT, anything
> > else in the
> > > > outer must be an error. But whereas RFC3168 dropped the packet if
> > > > there was CE in the outer, RFC4301 says forward it as Not-ECT
> > > > in all cases.
> > > >
> > > > RFC4301 gives no justification for this rather important
> > change from
> > > > RFC3168. I've spent the last hour or so trawling the
> > IPsec and tsvwg
> > > > archives and can't find any discussion on this point
> > either. Was it
> > > > discussed anywhere?
> > > >
> > > > This RFC4301 behaviour seems unsafe, but perhaps
> > understandable. It
> > > > all depends how the situation arose:
> > > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
> > > > then subsequently a congested router marked it CE, then
> > it's unsafe
> > > > to remove CE and forward Not-ECT - drop would have been safer.
> > > > * If something broken set CE on the outer even tho there was no
> > > > congestion, it's safe.
> > > >
> > > >
> > > >
> > > > Bob
> > > >
> > > > At 12:07 13/07/2009, Bob Briscoe wrote:
> > > > >David,
> > > > >
> > > > >We're getting somewhere (if we set aside (a) for now).
> > > > >
> > > > > > At 17:09 10/07/2009, Black_David@emc.com wrote:
> > > > >> > >Quick summary of the four changes proposed in the draft:
> > > > >> > >
> > > > >> > >a) Inner: Not-ECT, Outer: ECT(0)
> > > > >> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
> > > > >> > >b) Inner: ECT(0), Outer: ECT(1)
> > > > >> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
> > > > >> > >c) Inner: ECT(1), Outer: ECT(0)
> > > > >> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
> > > > >> > >d) Inner: CE, Outer: ECT(1)
> > > > >> > >         RFC 3168/4301: CE, Change: CE(!!!)
> > > > >> > >
> > > > >> > >If we were designing from scratch, all of these
> > changes would
> > > > >> > >be fine things to do.  We're not, and especially after the
> > > > >> > >discussion of how this impacts RFC 2003 gets added, this
> > > > >> > >draft will be proposing changes to some of the most widely
> > > > >> > >deployed mechanisms in the Internet.  To me, that strongly
> > > > >> > >suggests a conservative design approach - focus on
> > what really
> > > > >> > >needs to be changed, and resist them temptation to fix other
> > > > >> > >things that would be "nice to have".  In other words, IMHO
> > > > >> > >the appropriate criteria for "fixing it" is a very strong
> > > > >> > >cast that "it's broken".
> > > > >> > >
> > > > >> > >This is an area of principled disagreement on goals between
> > > > >> > >Bob and myself that list discussion is unlikely to resolve.
> > > > >> > >My view of the above four changes is that only b) makes it
> > > > >> > >over the "really needs to be changed" bar (and even then,
> > > > >> > >only with a *much* stronger statement of the PCN-based
> > > > >> > >rationale than is in the current draft, and Bob concurs
> > > > >> > >with strengthening the rationale).
> > > > >> > >
> > > > >> > >Turning to the other three changes, a) is clearly described
> > > > >> > >as "nice to have", enough said.
> > > > >> > >
> > > > >> > >For c) and d), I don't believe the optional alarms are
> > > > >> > >justified based on ECN (RFC 3168) alone.  A key observation
> > > > >> > >is that this draft does not make the < b) ECT(0) [Inner]
> > > > >> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
> > > > >> > >optional alarm.  What this means is that this draft not only
> > > > >> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
> > > > >> > >*assumes* that they are ordered in terms of priority,
> > > > >> > >severity, or something analogous.  That is unwarranted as
> > > > >> > >a general design assumption, IMHO.
> > > > >> > >
> > > > >> > >For this, reason, I think discussion of which
> > transitions are
> > > > >> > >optionally alarm-able in this fashion belongs in
> > the context of
> > > > >> > >a protocol that actually orders these codepoints, not in a
> > > > >> > >general document that would serve to "just constrain future
> > > > >> > >standards," borrowing Toby's words ;-).
> > > > >> > >
> > > > >> > >I could see discussion of optional alarms in an
> > Appendix that
> > > > >> > >discusses 3-valued PCN as an example, but not in
> > the main body
> > > > >> > >of the draft.  In the main body, it would be fine
> > to say that
> > > > >> > >specific uses of the ECN codepoints may assign meanings that
> > > > >> > >cause some of these transitions to be appropriate for alarms
> > > > >> > >(and point to the Appendix for a PCN example), but it would
> > > > >> > >not be ok to constrain the assignment of those meanings by
> > > > >> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
> > > > >> > >optionally alarm-able.
> > > > >> > >
> > > > >> > >Thanks,
> > > > >> > >--David
> > > > >> > >
> > > > >> > >
> > > > >> > > > -----Original Message-----
> > > > >> > > > From: tsvwg-bounces@ietf.org
> > [mailto:tsvwg-bounces@ietf.org]
> > > > >> > > > On Behalf Of Bob Briscoe
> > > > >> > > > Sent: Thursday, July 09, 2009 3:08 PM
> > > > >> > > > To: Black, David
> > > > >> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
> > > > >> > > > Subject: Re: [tsvwg] review of
> > > > >> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
> > > > >> > > >
> > > > >> > > > David,
> > > > >> > > >
> > > > >> > > > Bringing to the list as agreed. I'll reply in 2
> > > > parts. Part 1/2...
> > > > >> > > >
> > > > >> > > > Your comments get to the nub of the proposed egress
> > > > >> > protocol changes.
> > > > >> > > > Great... inline...
> > > > >> > > >
> > > > >> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
> > > > >> > > > >Bob,
> > > > >> > > > >
> > > > >> > > > >Sorry for the delay on this: things got entirely too
> > > > >> > busy around the
> > > > >> > > > >end of the quarter here.  Having re-read the draft from
> > > > >> > scratch, I
> > > > >> > > > >have a number of things to comment on ... I will be
> > > > in Stockholm,
> > > > >> > > > >and some of these items are going to need
> > further discussion
> > > > >> > > > >(private and/or public) in Stockholm.
> > > > >> > > > >
> > > > >> > > > >Starting with this table in Figure 4:
> > > > >> > > > >
> > > > >> > > >
> > > > >> >
> > > >
> > >+------------------+----------------------------------------------+
> > > > >> > > > >|  Incoming Inner  |             Incoming Outer Header
> > > > >> >          |
> > > > >> > > > >|      Header
> > > > >> > +---------+------------+------------+----------+
> > > > >> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
> > > > >> >    CE    |
> > > > >> > > >
> > > > >> >
> > > >
> > >+------------------+---------+------------+------------+----------+
> > > > >> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
> > > > >> > drop(!!!)|
> > > > >> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
> > > > >> >   CE     |
> > > > >> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
> > > > >> >   CE     |
> > > > >> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
> > > > >> >   CE     |
> > > > >> > > >
> > > > >> >
> > > >
> > >+------------------+---------+------------+------------+----------+
> > > > >> > > > >                    |                Outgoing Header
> > > > >> >           |
> > > > >> > > > >
> > > > >> > +----------------------------------------------+
> > > > >> > > > >
> > > > >> > > > >My initial focus is on what needs to change from RFC
> > > > >> > 3168/RFC 4301
> > > > >> > > > >behavior and why.  Given that these two RFCs line up, I
> > > > >> > think changes
> > > > >> > > > >need strong motivation.  So, here are the four
> > changes, and
> > > > >> > > > my thoughts:
> > > > >> > > > >
> > > > >> > > > >a) Inner: Not-ECT, Outer: ECT(0)
> > > > >> > > > >         3168/4301: drop (!!!)
> > > > >> > > > >         Change: Not-ECT (!!!)
> > > > >> > > > >         Rationale: "allow future standards actions
> > > > to use this
> > > > >> > > > >                 combination"
> > > > >> > > > >         My view: This ain't broke, why are you
> > fixing it?
> > > > >> > > >
> > > > >> > > > It is broke in the sense tunnel egresses unnecessarily
> > > > >> > drop harmless
> > > > >> > > > packets merely because they carry a "currently
> > unused" (CU)
> > > > >> > > > combination of codepoints. This is similar to the problem
> > > > >> > ECN faced
> > > > >> > > > with firewalls dropping packets unless they were Not-ECT.
> > > > >> > > >
> > > > >> > > > (a) wasn't originally part of the changes I was
> > > > >> > proposing. But Anil
> > > > >> > > > Agawaal convinced me it is worth doing. Reasoning: If we
> > > > >> > are making a
> > > > >> > > > change to decap anyway, let's get it as right as
> > > > possible. If we
> > > > >> > > > decide not to make any other changes to decap, I wouldn't
> > > > >> > fight for
> > > > >> > > > this one alone.
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > >b) Inner: ECT(0), Outer: ECT(1)
> > > > >> > > > >         3168/4301: ECT(0)
> > > > >> > > > >         Change: ECT(1)
> > > > >> > > > >         Rationale: Treat ECT(1) as more severe
> > congestion
> > > > >> > > > than ECT(0).
> > > > >> > > > >         My view: Longer discussion below.
> > > > >> > > > >
> > > > >> > > > >c) Inner: ECT(1), Outer: ECT(0)
> > > > >> > > > >         3168/4301: ECT(1)
> > > > >> > > > >         Change: ECT(1) (!!!)
> > > > >> > > > >         Rationale: "no currently legal transition" can
> > > > >> > > > produce this combo.
> > > > >> > > > >         My view: Not a good idea, among other reasons
> > > > >> > > > because it's at
> > > > >> > > > >                 odds with your "future
> > standards action"
> > > > >> > > > explanation for
> > > > >> > > > >                 the first change.  I'd remove
> > this change
> > > > >> > > > and instead deal
> > > > >> > > > >                 with the possibility of
> > raising an alarm in
> > > > >> > > > a PCN-specific
> > > > >> > > > >                 draft rather than declaring
> > this to be an
> > > > >> > > > alarm-able combo
> > > > >> > > > >                 in general.
> > > > >> > > > >
> > > > >> > > > >d) Inner: CE, Outer: ECT(1)
> > > > >> > > > >         3168/4301: CE
> > > > >> > > > >         Change: CE (!!!)
> > > > >> > > > >         Rationale & My view: <same as previous change>
> > > > >> > > >
> > > > >> > > > c) & d) aren't PCN specific.
> > > > >> > > >
> > > > >> > > > Reasoning for adding alarms (also said this eariler
> > > > on list in
> > > > >> > > > conversation with Gorry):
> > > > >> > > > If security folks have concerns over possible attacks
> > > > >> > from "currently
> > > > >> > > > unused" combinations, I want to give them a way
> > to manage the
> > > > >> > > > perceived threat without just dropping packets. RFC3168
> > > > >> > recommended
> > > > >> > > > alarms for CU combinations. In these cases, I believe
> > > > >> > RFC3168 simply
> > > > >> > > > overlooked that these are also CU combinations.
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > >So, I would remove a) as not defensible (no good reason
> > > > >> > for change),
> > > > >> > > > >and I would push the discussion of adding alarms for c)
> > > > >> > and d) into
> > > > >> > > > >a PCN-specific document.
> > > > >> > > >
> > > > >> > > > c) & d) are not to do with PCN.
> > > > >> > > >
> > > > >> > > > My motivation is: While making change (b) to enable
> > > > PCN, lets do
> > > > >> > > > (a,c,d) to make the IP layer as clean as possible, to
> > > > >> > ensure easier
> > > > >> > > > future evolvability.
> > > > >> > > >
> > > > >> > > > I don't see the issue. It says SHOULD log and MAY
> > > > alarm. So the
> > > > >> > > > operator doesn't have to do anything.
> > > > >> > > >
> > > > >> > > > >The c) alarm in particular is only
> > > > >> > > > >meaningful if PCN or something like it is in use.
> > > > >> > > >
> > > > >> > > > No. The combination in c) always was CU, and it's still
> > > > >> > CU with PCN.
> > > > >> > > > None of the proposed changes are PCN-specific.
> > > > >> > > >
> > > > >> > > > Even (b) is intended to also be useful for any 3-level
> > > > >> > ECN while not
> > > > >> > > > precluding 2-level ECN and the nonce.
> > > > >> > > >
> > > > >> > > > NB:  3 level = 2 'congested' levels + 1 level not
> > > > >> > congested, e.g. VCP
> > > > >> > > > or PCN which were both mentioned in RFC4774. This doesn't
> > > > >> > count the
> > > > >> > > > not-ECN-capable state, which makes 4 states.
> > > > >> > > >
> > > > >> > > > >I don't view d)
> > > > >> > > > >as deserving of an alarm in general, as even though
> > > > something has
> > > > >> > > > >happened, no damage has been done - this is in
> > line with the
> > > > >> > > > >"be liberal in what you accept" principle of
> > > > protocol design.
> > > > >> > > >
> > > > >> > > > That's why I'm proposing an optional alarm. Optional
> > > > alarms *are*
> > > > >> > > > liberal, but they allow operators a way to address
> > > > >> > security concerns
> > > > >> > > > about 'known unknowns' [Rumsfelt], while still
> > allowing new
> > > > >> > > > behaviours to be introduced.
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > >That leaves b), which I think is the key desired
> > > > change, because
> > > > >> > > > >it makes a third ECN code point usable through tunnels.
> > > > >> > > >
> > > > >> > > > Yes - exactly.
> > > > >> > > >
> > > > >> > > > >Much as
> > > > >> > > > >this might be useful for PCN, it may be too late to
> > > > do this.  The
> > > > >> > > > >problem starts from the need to say that this draft
> > > > >> > updates RFC 4301.
> > > > >> > > > >The result will be a mix of 4301 and updated-4301
> > > > implementations
> > > > >> > > > >that render the third state difficult to use in
> > > > practice.  The
> > > > >> > > > >problem's even worse for non-IPsec tunnels.
> > > > >> > > >
> > > > >> > > > The idea is that an operator wanting to implement
> > > > PCN will have a
> > > > >> > > > spec they can point to in their invitation to
> > > > tender. At the same
> > > > >> > > > time this spec both UPDATEs and is backward compatible
> > > > >> > with RFC3168 &
> > > > >> > > > RFC4301 because it only changes CU stuff at egress and
> > > > >> > the change at
> > > > >> > > > the ingress leads to the same output at the egress.
> > > > >> > Therefore vendors
> > > > >> > > > can update 3168 & 4301 tunnels lazily - only when their
> > > > >> > customers ask
> > > > >> > > > for the new features (whether for PCN or for something
> > > > >> > else like VCP).
> > > > >> > > >
> > > > >> > > > >I also read Appendix D
> > > > >> > > > >as saying that PCN is going the two-state
> > route, not three.
> > > > >> > > >
> > > > >> > > > That's not how it should be read - I'll clarify it. The
> > > > >> > PCN w-g wants
> > > > >> > > > 3 states, but can't have them while ecn-tunnel is
> > > > not proposed
> > > > >> > > > standard. Therefore, it has decided to define a 2-state
> > > > >> > baseline and
> > > > >> > > > a number of experimental schemes to work round this
> > > > tunnel issue
> > > > >> > > > (plus a comparison doc about them all). It has agreed
> > > > >> > that eventually
> > > > >> > > > just one experimental scheme will be chosen for
> > > > standards track:
> > > > >> > > > - 3-in-1 (the simplest, which depends on (b) of
> > ecn-tunnel)
> > > > >> > > > - 3-state (which uses an extra DSCP)
> > > > >> > > > - psdm (which requires out of band state)
> > > > >> > > > - sm (single marking gets 3 functions from 2
> > > > encoding states but
> > > > >> > > > admits admission control & termination aren't
> > very precise in
> > > > >> > > > all cases)
> > > > >> > > >
> > > > >> > > > ___
> > > > >> > > > See part 2/2 for responses to your comments on
> > specific parts
> > > > >> > > > of the test.
> > > > >> > > >
> > > > >> > > > Thanks again for the review
> > > > >> > > >
> > > > >> > > >
> > > > >> > > > Bob
> > > > >> > > >
> > > > >> > > >
> > > > >> > > >
> > > > >> > > >
> > > > ________________________________________________________________
> > > > >> > > > Bob Briscoe,               Networks Research Centre,
> > > > BT Research
> > > > >> > > >
> > > > >> > > >
> > > > >> > > >
> > > > >> >
> > > > >> >
> > ________________________________________________________________
> > > > >> > Bob Briscoe,               Networks Research Centre,
> > BT Research
> > > > >> >
> > > > >> >
> > > > >> >
> > > > >
> > > > >________________________________________________________________
> > > > >Bob Briscoe,               Networks Research Centre, BT Research
> > > >
> > > > ________________________________________________________________
> > > > Bob Briscoe,               Networks Research Centre, BT Research
> > > >
> > > >
> > > >
> >
> > ________________________________________________________________
> > Bob Briscoe,               Networks Research Centre, BT Research
> >
> >
> >

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research 


From flefauch@cisco.com  Mon Jul 27 01:00:28 2009
Return-Path: <flefauch@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id B295628C14D for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:00:28 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -10.099
X-Spam-Level: 
X-Spam-Status: No, score=-10.099 tagged_above=-999 required=5 tests=[AWL=0.500, BAYES_00=-2.599, RCVD_IN_DNSWL_HI=-8]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id Ulv+v6eFiksI for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:00:27 -0700 (PDT)
Received: from ams-iport-1.cisco.com (ams-iport-1.cisco.com [144.254.224.140]) by core3.amsl.com (Postfix) with ESMTP id 6335528C108 for <tsvwg@ietf.org>; Mon, 27 Jul 2009 01:00:27 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AkkAAGv7bEqQ/uCLe2dsb2JhbACZfwEBFiQGnh6IKI1fBYQN
X-IronPort-AV: E=Sophos;i="4.43,275,1246838400"; d="scan'208";a="45879348"
Received: from ams-dkim-2.cisco.com ([144.254.224.139]) by ams-iport-1.cisco.com with ESMTP; 27 Jul 2009 08:00:27 +0000
Received: from ams-core-1.cisco.com (ams-core-1.cisco.com [144.254.224.150]) by ams-dkim-2.cisco.com (8.12.11/8.12.11) with ESMTP id n6R80QKJ003679;  Mon, 27 Jul 2009 10:00:26 +0200
Received: from xbh-ams-102.cisco.com (xbh-ams-102.cisco.com [144.254.73.132]) by ams-core-1.cisco.com (8.13.8/8.14.3) with ESMTP id n6R80RKv009474; Mon, 27 Jul 2009 08:00:27 GMT
Received: from xfe-ams-331.emea.cisco.com ([144.254.231.72]) by xbh-ams-102.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 10:00:27 +0200
Received: from dhcp-15a6.meeting.ietf.org ([10.61.69.207]) by xfe-ams-331.emea.cisco.com with Microsoft SMTPSVC(6.0.3790.1830);  Mon, 27 Jul 2009 10:00:26 +0200
Message-Id: <B988E5CF-DBF4-4B93-B58F-B521A108570B@cisco.com>
From: Francois Le Faucheur IMAP <flefauch@cisco.com>
To: "James M. Polk" <jmpolk@cisco.com>
In-Reply-To: <XFE-SJC-212AJNACcKl00009d8b@xfe-sjc-212.amer.cisco.com>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (Apple Message framework v935.3)
Date: Mon, 27 Jul 2009 10:00:25 +0200
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com> <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com> <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com> <XFE-SJC-212AJNACcKl00009d8b@xfe-sjc-212.amer.cisco.com>
X-Mailer: Apple Mail (2.935.3)
X-OriginalArrivalTime: 27 Jul 2009 08:00:27.0002 (UTC) FILETIME=[4D64EDA0:01CA0E90]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=1524; t=1248681626; x=1249545626; c=relaxed/simple; s=amsdkim2001; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=flefauch@cisco.com; z=From:=20Francois=20Le=20Faucheur=20IMAP=20<flefauch@cisco. com> |Subject:=20Re=3A=20[tsvwg]=20New=20rev=20of=20IntServ=20MU LTI_TSPEC |Sender:=20; bh=hFGSNw2Zyh/2Rq7S4051cqah1r0CeAV63Fmlz+gx16I=; b=oLTksNq6qjj6Q3B0OoU8remwWjg/0kNO1fBMV+NT1ydW+qGIIAISSo9sbH 0oBbSy1bOOwUjGvUc+z4eg75LsPHJfYV7rn+rFJtgJXd0kgMk3ykN22n2Iju AAm3Qux79z;
Authentication-Results: ams-dkim-2; header.From=flefauch@cisco.com; dkim=pass ( sig from cisco.com/amsdkim2001 verified; ); 
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 27 Jul 2009 08:00:28 -0000

James, Lars,

On 27 Jul 2009, at 01:02, James M. Polk wrote:

> At 05:54 PM 7/26/2009, Lars Eggert wrote:
>> Hi,
>>
>> On 2009-7-26, at 18:00, Francois Le Faucheur IMAP wrote:
>>> If you are not protected against malicious RSVP attackers, there are
>>> lots of very bad things you have to worry about (probably more
>>> worrisome than the cramming of many TSPECs in a message). For
>>> starters, the attacker can simply reserve all remaining bandwidth to
>>> all destinations. Hence the recommendation to implement neighbor
>>> authentication, which will protect you against malicious attackers.
>>
>> OK, but then you need to require use of RFC2747, which at the moment
>> isn't even mentioned in the draft.
>
> sorry for missing this.
>
> I'll get it in the next rev.

FYI, after going through similar discussions as part of the AD review  
of draft-ietf-tsvwg-rsvp-l3vpn, we agreed to include the following  
text in the Security Considerations section:

" To ensure the integrity of RSVP, the RSVP Authentication mechanisms  
defined in [RFC2747] and [RFC3097] SHOULD be used. Those protect RSVP  
message integrity hop-by-hop and provide node authentication as well  
as replay protection, thereby protecting against corruption and  
spoofing of RSVP messages."

I suggest you include same or similar text.

BTW, there may be more text from the security considerations of draft- 
ietf-tsvwg-rsvp-l3vpn that may be worth considering.

Francois


>
> James
>
>
>> Lars
>>
>


From lars.eggert@nokia.com  Mon Jul 27 01:31:40 2009
Return-Path: <lars.eggert@nokia.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 3118F28C13F for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:31:40 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.551
X-Spam-Level: 
X-Spam-Status: No, score=-2.551 tagged_above=-999 required=5 tests=[AWL=0.048,  BAYES_00=-2.599]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id CqNSEfUDqELz for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:31:39 -0700 (PDT)
Received: from mail.fit.nokia.com (mail.fit.nokia.com [195.148.124.195]) by core3.amsl.com (Postfix) with ESMTP id 3464D3A68CF for <tsvwg@ietf.org>; Mon, 27 Jul 2009 01:31:39 -0700 (PDT)
Received: from [IPv6:2001:df8::80:225:ff:fe45:eccf] ([IPv6:2001:df8:0:80:225:ff:fe45:eccf]) (authenticated bits=0) by mail.fit.nokia.com (8.14.3/8.14.3) with ESMTP id n6R8VRgc063120 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=NOT); Mon, 27 Jul 2009 11:31:28 +0300 (EEST) (envelope-from lars.eggert@nokia.com)
Message-Id: <FBA51016-DCA9-4DA2-AAFE-26390AD33A7F@nokia.com>
From: Lars Eggert <lars.eggert@nokia.com>
To: Francois Le Faucheur IMAP <flefauch@cisco.com>
In-Reply-To: <B988E5CF-DBF4-4B93-B58F-B521A108570B@cisco.com>
Content-Type: multipart/signed; boundary=Apple-Mail-95--206657258; micalg=sha1; protocol="application/pkcs7-signature"
Mime-Version: 1.0 (Apple Message framework v935.3)
Date: Mon, 27 Jul 2009 10:31:22 +0200
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com> <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com> <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com> <XFE-SJC-212AJNACcKl00009d8b@xfe-sjc-212.amer.cisco.com> <B988E5CF-DBF4-4B93-B58F-B521A108570B@cisco.com>
X-Mailer: Apple Mail (2.935.3)
X-Greylist: Sender succeeded SMTP AUTH, not delayed by milter-greylist-4.2.2 (mail.fit.nokia.com [IPv6:2001:2060:40:1::123]); Mon, 27 Jul 2009 11:31:29 +0300 (EEST)
Cc: "James M. Polk" <jmpolk@cisco.com>, tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 27 Jul 2009 08:31:40 -0000

--Apple-Mail-95--206657258
Content-Type: text/plain;
	charset=US-ASCII;
	format=flowed;
	delsp=yes
Content-Transfer-Encoding: 7bit

Hi,

the text below looks OK, but I really believe this needs to be a MUST  
here, because of the additional attack angle.

Lars

On 2009-7-27, at 10:00, Francois Le Faucheur IMAP wrote:
> FYI, after going through similar discussions as part of the AD review
> of draft-ietf-tsvwg-rsvp-l3vpn, we agreed to include the following
> text in the Security Considerations section:
>
> " To ensure the integrity of RSVP, the RSVP Authentication mechanisms
> defined in [RFC2747] and [RFC3097] SHOULD be used. Those protect RSVP
> message integrity hop-by-hop and provide node authentication as well
> as replay protection, thereby protecting against corruption and
> spoofing of RSVP messages."
>
> I suggest you include same or similar text.
>
> BTW, there may be more text from the security considerations of draft-
> ietf-tsvwg-rsvp-l3vpn that may be worth considering.


--Apple-Mail-95--206657258
Content-Disposition: attachment;
	filename=smime.p7s
Content-Type: application/pkcs7-signature;
	name=smime.p7s
Content-Transfer-Encoding: base64
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==

--Apple-Mail-95--206657258--

From jmpolk@cisco.com  Mon Jul 27 01:57:42 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 7E8573A6824 for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:57:42 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.431
X-Spam-Level: 
X-Spam-Status: No, score=-6.431 tagged_above=-999 required=5 tests=[AWL=0.168,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 7xQx7tdN-wnB for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:57:41 -0700 (PDT)
Received: from sj-iport-3.cisco.com (sj-iport-3.cisco.com [171.71.176.72]) by core3.amsl.com (Postfix) with ESMTP id 83A2328C1BD for <tsvwg@ietf.org>; Mon, 27 Jul 2009 01:57:41 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AlUFAH8IbUqrR7MV/2dsb2JhbACIR7AGiCiNZAWEDQ
X-IronPort-AV: E=Sophos;i="4.43,275,1246838400"; d="scan'208";a="179595218"
Received: from sj-dkim-1.cisco.com ([171.71.179.21]) by sj-iport-3.cisco.com with ESMTP; 27 Jul 2009 08:57:42 +0000
Received: from sj-core-1.cisco.com (sj-core-1.cisco.com [171.71.177.237]) by sj-dkim-1.cisco.com (8.12.11/8.12.11) with ESMTP id n6R8vgJB000669;  Mon, 27 Jul 2009 01:57:42 -0700
Received: from xbh-sjc-231.amer.cisco.com (xbh-sjc-231.cisco.com [128.107.191.100]) by sj-core-1.cisco.com (8.13.8/8.14.3) with ESMTP id n6R8vgdB002135; Mon, 27 Jul 2009 08:57:42 GMT
Received: from xfe-sjc-212.amer.cisco.com ([171.70.151.187]) by xbh-sjc-231.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 01:57:42 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.4.153]) by xfe-sjc-212.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 01:57:41 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 27 Jul 2009 03:57:37 -0500
To: Francois Le Faucheur IMAP <flefauch@cisco.com>
From: "James M. Polk" <jmpolk@cisco.com>
In-Reply-To: <B988E5CF-DBF4-4B93-B58F-B521A108570B@cisco.com>
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com> <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com> <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com> <XFE-SJC-212AJNACcKl00009d8b@xfe-sjc-212.amer.cisco.com> <B988E5CF-DBF4-4B93-B58F-B521A108570B@cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-212AmX3qrB000009e4c@xfe-sjc-212.amer.cisco.com>
X-OriginalArrivalTime: 27 Jul 2009 08:57:41.0920 (UTC) FILETIME=[4CC3D200:01CA0E98]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=1636; t=1248685062; x=1249549062; c=relaxed/simple; s=sjdkim1004; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20Re=3A=20[tsvwg]=20New=20rev=20of=20IntServ=20MU LTI_TSPEC |Sender:=20; bh=zlRed9XQKahlDZU8ejgSPapzhV/JyeDaDTYwFrsZwB4=; b=gjTCCpSZ57SHxdQ9didPboY/V7d713sCeXXyWyE9c/AunKNo4UP+kc/yW8 xoiYjvjA253mNdTn1CH24DPJocnkrf3WOVk0iritKv208tV/knr/EegGfyar qSwFa74mFsRaaVzO/NHB0JkNcOC5qwMfm8WFBP66gZi4mwP4ACdVw=;
Authentication-Results: sj-dkim-1; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim1004 verified; ); 
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 27 Jul 2009 08:57:42 -0000

At 03:00 AM 7/27/2009, Francois Le Faucheur IMAP wrote:
>James, Lars,
>
>On 27 Jul 2009, at 01:02, James M. Polk wrote:
>
>>At 05:54 PM 7/26/2009, Lars Eggert wrote:
>>>Hi,
>>>
>>>On 2009-7-26, at 18:00, Francois Le Faucheur IMAP wrote:
>>>>If you are not protected against malicious RSVP attackers, there are
>>>>lots of very bad things you have to worry about (probably more
>>>>worrisome than the cramming of many TSPECs in a message). For
>>>>starters, the attacker can simply reserve all remaining bandwidth to
>>>>all destinations. Hence the recommendation to implement neighbor
>>>>authentication, which will protect you against malicious attackers.
>>>
>>>OK, but then you need to require use of RFC2747, which at the moment
>>>isn't even mentioned in the draft.
>>
>>sorry for missing this.
>>
>>I'll get it in the next rev.
>
>FYI, after going through similar discussions as part of the AD review
>of draft-ietf-tsvwg-rsvp-l3vpn, we agreed to include the following
>text in the Security Considerations section:
>
>" To ensure the integrity of RSVP, the RSVP Authentication mechanisms
>defined in [RFC2747] and [RFC3097] SHOULD be used. Those protect RSVP
>message integrity hop-by-hop and provide node authentication as well
>as replay protection, thereby protecting against corruption and
>spoofing of RSVP messages."
>
>I suggest you include same or similar text.

this is good, I will -- thanks Francois


>BTW, there may be more text from the security considerations of 
>draft- ietf-tsvwg-rsvp-l3vpn that may be worth considering.
>
>Francois
>
>
>>
>>James
>>
>>
>>>Lars


From jmpolk@cisco.com  Mon Jul 27 01:58:28 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E331928C25B for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:58:28 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.44
X-Spam-Level: 
X-Spam-Status: No, score=-6.44 tagged_above=-999 required=5 tests=[AWL=0.159,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id JsIDKuAymt7o for <tsvwg@core3.amsl.com>; Mon, 27 Jul 2009 01:58:28 -0700 (PDT)
Received: from sj-iport-6.cisco.com (sj-iport-6.cisco.com [171.71.176.117]) by core3.amsl.com (Postfix) with ESMTP id 324A528C259 for <tsvwg@ietf.org>; Mon, 27 Jul 2009 01:58:28 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AlUFALsIbUqrR7O6/2dsb2JhbACIR7AJiCiNZAWEDQ
X-IronPort-AV: E=Sophos;i="4.43,275,1246838400"; d="scan'208";a="354565163"
Received: from sj-dkim-2.cisco.com ([171.71.179.186]) by sj-iport-6.cisco.com with ESMTP; 27 Jul 2009 08:58:29 +0000
Received: from sj-core-2.cisco.com (sj-core-2.cisco.com [171.71.177.254]) by sj-dkim-2.cisco.com (8.12.11/8.12.11) with ESMTP id n6R8wToX020754;  Mon, 27 Jul 2009 01:58:29 -0700
Received: from xbh-sjc-221.amer.cisco.com (xbh-sjc-221.cisco.com [128.107.191.63]) by sj-core-2.cisco.com (8.13.8/8.14.3) with ESMTP id n6R8wT45018804; Mon, 27 Jul 2009 08:58:29 GMT
Received: from xfe-sjc-212.amer.cisco.com ([171.70.151.187]) by xbh-sjc-221.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 01:58:29 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.4.153]) by xfe-sjc-212.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Mon, 27 Jul 2009 01:58:28 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Mon, 27 Jul 2009 03:58:24 -0500
To: Lars Eggert <lars.eggert@nokia.com>, Francois Le Faucheur IMAP <flefauch@cisco.com>
From: "James M. Polk" <jmpolk@cisco.com>
In-Reply-To: <FBA51016-DCA9-4DA2-AAFE-26390AD33A7F@nokia.com>
References: <XFE-SJC-212USp3sJWG000080ba@xfe-sjc-212.amer.cisco.com> <3C1F86AD-D81E-4816-9D5F-CC8065B17545@nokia.com> <260B5715-2DA4-47FE-88E3-158F94FDA5FF@cisco.com> <57C993B8-B0AE-4C46-B54D-C6F883D0844D@nokia.com> <XFE-SJC-212AJNACcKl00009d8b@xfe-sjc-212.amer.cisco.com> <B988E5CF-DBF4-4B93-B58F-B521A108570B@cisco.com> <FBA51016-DCA9-4DA2-AAFE-26390AD33A7F@nokia.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-212Xn2Pi1Y700009e4d@xfe-sjc-212.amer.cisco.com>
X-OriginalArrivalTime: 27 Jul 2009 08:58:28.0607 (UTC) FILETIME=[6897B0F0:01CA0E98]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=990; t=1248685109; x=1249549109; c=relaxed/simple; s=sjdkim2002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20Re=3A=20[tsvwg]=20New=20rev=20of=20IntServ=20MU LTI_TSPEC |Sender:=20; bh=hwpOMiV3hd72EOUueeYeXf3MXnPvk8BrKoFyiusmdV4=; b=0fX8/l6aXdLcsqlqQlT0GBXN9j63NH3lSOjJadwR0+jyfgHbPvxbG8twHv zsSN4d3cFzkORGMP43ts7JGEOsS3k1y1nb3YVvf1lvDFg+q/bUBIJdTzj7BR SkHMoT5NbV;
Authentication-Results: sj-dkim-2; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim2002 verified; ); 
Cc: tsvwg <tsvwg@ietf.org>
Subject: Re: [tsvwg] New rev of IntServ MULTI_TSPEC
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 27 Jul 2009 08:58:29 -0000

At 03:31 AM 7/27/2009, Lars Eggert wrote:
>Hi,
>
>the text below looks OK, but I really believe this needs to be a MUST
>here, because of the additional attack angle.

I can easily upgrade this to a MUST.


>Lars
>
>On 2009-7-27, at 10:00, Francois Le Faucheur IMAP wrote:
>>FYI, after going through similar discussions as part of the AD review
>>of draft-ietf-tsvwg-rsvp-l3vpn, we agreed to include the following
>>text in the Security Considerations section:
>>
>>" To ensure the integrity of RSVP, the RSVP Authentication mechanisms
>>defined in [RFC2747] and [RFC3097] SHOULD be used. Those protect RSVP
>>message integrity hop-by-hop and provide node authentication as well
>>as replay protection, thereby protecting against corruption and
>>spoofing of RSVP messages."
>>
>>I suggest you include same or similar text.
>>
>>BTW, there may be more text from the security considerations of draft-
>>ietf-tsvwg-rsvp-l3vpn that may be worth considering.
>
>
>


From root@core3.amsl.com  Mon Jul 27 15:00:01 2009
Return-Path: <root@core3.amsl.com>
X-Original-To: tsvwg@ietf.org
Delivered-To: tsvwg@core3.amsl.com
Received: by core3.amsl.com (Postfix, from userid 0) id 322043A68DC; Mon, 27 Jul 2009 15:00:00 -0700 (PDT)
From: Internet-Drafts@ietf.org
To: i-d-announce@ietf.org
Content-Type: Multipart/Mixed; Boundary="NextPart"
Mime-Version: 1.0
Message-Id: <20090727220001.322043A68DC@core3.amsl.com>
Date: Mon, 27 Jul 2009 15:00:01 -0700 (PDT)
Cc: tsvwg@ietf.org
Subject: [tsvwg] I-D Action:draft-ietf-tsvwg-ecn-tunnel-03.txt
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Mon, 27 Jul 2009 22:00:01 -0000

--NextPart

A New Internet-Draft is available from the on-line Internet-Drafts directories.
This draft is a work item of the Transport Area Working Group Working Group of the IETF.


	Title           : Tunnelling of Explicit Congestion Notification
	Author(s)       : B. Briscoe
	Filename        : draft-ietf-tsvwg-ecn-tunnel-03.txt
	Pages           : 40
	Date            : 2009-07-27

This document redefines how the explicit congestion notification
(ECN) field of the IP header should be constructed on entry to and
exit from any IP in IP tunnel.  On encapsulation it updates RFC3168
to bring all IP in IP tunnels (v4 or v6) into line with RFC4301 IPsec
ECN processing.  On decapsulation it updates both RFC3168 and RFC4301
to add new behaviours for previously unused combinations of inner and
outer header.  The new rules propagate the ECN field whether it is
used to signal one or two severity levels of congestion, whereas
before they propagated only one.  Tunnel endpoints can be updated in
any order without affecting pre-existing uses of the ECN field
(backward compatible).  Nonetheless, operators wanting to support two
severity levels (e.g. for pre-congestion notification--PCN) can
require compliance with this new specification.  A thorough analysis
of the reasoning for these changes and the implications is included.

A URL for this Internet-Draft is:
http://www.ietf.org/internet-drafts/draft-ietf-tsvwg-ecn-tunnel-03.txt

Internet-Drafts are also available by anonymous FTP at:
ftp://ftp.ietf.org/internet-drafts/

Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.

--NextPart
Content-Type: Message/External-body;
	name="draft-ietf-tsvwg-ecn-tunnel-03.txt";
	site="ftp.ietf.org";
	access-type="anon-ftp";
	directory="internet-drafts"

Content-Type: text/plain
Content-ID: <2009-07-27145942.I-D@ietf.org>


--NextPart--

From jmpolk@cisco.com  Tue Jul 28 04:19:03 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E9A703A6847 for <tsvwg@core3.amsl.com>; Tue, 28 Jul 2009 04:19:03 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.499
X-Spam-Level: 
X-Spam-Status: No, score=-6.499 tagged_above=-999 required=5 tests=[AWL=0.100,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 8cP+1mvKLlgV for <tsvwg@core3.amsl.com>; Tue, 28 Jul 2009 04:19:03 -0700 (PDT)
Received: from sj-iport-6.cisco.com (sj-iport-6.cisco.com [171.71.176.117]) by core3.amsl.com (Postfix) with ESMTP id 206F53A6DC4 for <tsvwg@ietf.org>; Tue, 28 Jul 2009 04:19:02 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: ApIGAAt7bkqrR7PD/2dsb2JhbACIR7IXiCePSwWEDA
X-IronPort-AV: E=Sophos;i="4.43,282,1246838400"; d="scan'208";a="355417847"
Received: from sj-dkim-3.cisco.com ([171.71.179.195]) by sj-iport-6.cisco.com with ESMTP; 28 Jul 2009 11:19:02 +0000
Received: from sj-core-5.cisco.com (sj-core-5.cisco.com [171.71.177.238]) by sj-dkim-3.cisco.com (8.12.11/8.12.11) with ESMTP id n6SBJ2vV024883 for <tsvwg@ietf.org>; Tue, 28 Jul 2009 04:19:02 -0700
Received: from xbh-sjc-221.amer.cisco.com (xbh-sjc-221.cisco.com [128.107.191.63]) by sj-core-5.cisco.com (8.13.8/8.14.3) with ESMTP id n6SBJ2qU013664 for <tsvwg@ietf.org>; Tue, 28 Jul 2009 11:19:02 GMT
Received: from xfe-sjc-211.amer.cisco.com ([171.70.151.174]) by xbh-sjc-221.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Tue, 28 Jul 2009 04:19:01 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.11.121]) by xfe-sjc-211.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Tue, 28 Jul 2009 04:18:59 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Tue, 28 Jul 2009 06:18:52 -0500
To: tsvwg <tsvwg@ietf.org>
From: "James M. Polk" <jmpolk@cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-211Ei8TywMf000051f5@xfe-sjc-211.amer.cisco.com>
X-OriginalArrivalTime: 28 Jul 2009 11:19:01.0468 (UTC) FILETIME=[3561E9C0:01CA0F75]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=148; t=1248779942; x=1249643942; c=relaxed/simple; s=sjdkim3002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20TSVWG=20call=20for=20note=20takers |Sender:=20; bh=OsOJV3d1umrgP7MuCwHmga2PazVMc5iPh1yagVL0CTY=; b=p7INa7M65mE5dHJnQl/bmwgjCcBD0d2gniXcLcifUdnFeCc7jSCQZ/AXza sGgxoV16byAI8JZrR6EjV7fcD2HOLWN8nmAh4TBwmmA9rp183l2E5X22cOIv XUikB/Rt3F;
Authentication-Results: sj-dkim-3; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim3002 verified; ); 
Subject: [tsvwg] TSVWG call for note takers
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 28 Jul 2009 11:19:04 -0000

TSVWG

We chairs are looking for a couple of note takers during the meeting 
on Thursday.

Does anyone want to volunteer now?

James
Gorry


From jmpolk@cisco.com  Tue Jul 28 16:28:05 2009
Return-Path: <jmpolk@cisco.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id E7A0E3A685D for <tsvwg@core3.amsl.com>; Tue, 28 Jul 2009 16:28:05 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.517
X-Spam-Level: 
X-Spam-Status: No, score=-6.517 tagged_above=-999 required=5 tests=[AWL=0.082,  BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id VKuw7dSaadAS for <tsvwg@core3.amsl.com>; Tue, 28 Jul 2009 16:28:05 -0700 (PDT)
Received: from sj-iport-6.cisco.com (sj-iport-6.cisco.com [171.71.176.117]) by core3.amsl.com (Postfix) with ESMTP id 33DC23A6840 for <tsvwg@ietf.org>; Tue, 28 Jul 2009 16:28:05 -0700 (PDT)
X-IronPort-Anti-Spam-Filtered: true
X-IronPort-Anti-Spam-Result: AkcFACMmb0qrR7PE/2dsb2JhbACIR7Q/iCeQKAWEEIFN
X-IronPort-AV: E=Sophos;i="4.43,285,1246838400"; d="scan'208";a="355887170"
Received: from sj-dkim-4.cisco.com ([171.71.179.196]) by sj-iport-6.cisco.com with ESMTP; 28 Jul 2009 23:28:06 +0000
Received: from sj-core-5.cisco.com (sj-core-5.cisco.com [171.71.177.238]) by sj-dkim-4.cisco.com (8.12.11/8.12.11) with ESMTP id n6SNS64T015902 for <tsvwg@ietf.org>; Tue, 28 Jul 2009 16:28:06 -0700
Received: from xbh-sjc-211.amer.cisco.com (xbh-sjc-211.cisco.com [171.70.151.144]) by sj-core-5.cisco.com (8.13.8/8.14.3) with ESMTP id n6SNS689019967 for <tsvwg@ietf.org>; Tue, 28 Jul 2009 23:28:06 GMT
Received: from xfe-sjc-211.amer.cisco.com ([171.70.151.174]) by xbh-sjc-211.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Tue, 28 Jul 2009 16:28:06 -0700
Received: from jmpolk-wxp01.cisco.com ([10.89.5.89]) by xfe-sjc-211.amer.cisco.com with Microsoft SMTPSVC(6.0.3790.3959);  Tue, 28 Jul 2009 16:28:05 -0700
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Tue, 28 Jul 2009 18:27:55 -0500
To: tsvwg <tsvwg@ietf.org>
From: "James M. Polk" <jmpolk@cisco.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Message-ID: <XFE-SJC-2114U135Tcf00005383@xfe-sjc-211.amer.cisco.com>
X-OriginalArrivalTime: 28 Jul 2009 23:28:06.0502 (UTC) FILETIME=[0F746060:01CA0FDB]
DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; l=579; t=1248823686; x=1249687686; c=relaxed/simple; s=sjdkim4002; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jmpolk@cisco.com; z=From:=20=22James=20M.=20Polk=22=20<jmpolk@cisco.com> |Subject:=20TSVWG=20Agenda=20item=20change |Sender:=20; bh=1GNoKNi20u3KVPAwKME9+H95+lVeTp0DwrpN/MxBmkE=; b=s4TWWT/wWPKlYpm8eNRoVsRpKw6tl9VqS/PMTqO0Ad7xNHI+cwm0NJCR+L IMsDOLW+ewh75oZi4intRm72nytHHmzvSFBYQWYFohKsrioiqcVwIFp49ADv 0vvLY/TjyT;
Authentication-Results: sj-dkim-4; header.From=jmpolk@cisco.com; dkim=pass ( sig from cisco.com/sjdkim4002 verified; ); 
Subject: [tsvwg] TSVWG Agenda item change
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Tue, 28 Jul 2009 23:28:06 -0000

TSVWG

The chairs have added an SCTP update to the agenda, much like we had 
in IETF74 - in which Michael will go over all the currently 
progressing SCTP documents and provide status (mods, issues, dependencies).

http://www.ietf.org/proceedings/75/agenda/tsvwg.txt

This is being slotted right after the WG chairs' presentation (i.e., 
2nd in order), mostly due to a conflict with HTTPBIS WG that has an 
effort their involving HTTP over SCTP, with the SCTP authors wanting 
to attend that WG also.  We have accommodated this request.

James
Gorry
TSVWG chairs


From mohammad.arifb@gmail.com  Wed Jul 29 02:38:58 2009
Return-Path: <mohammad.arifb@gmail.com>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 48BE93A69EC for <tsvwg@core3.amsl.com>; Wed, 29 Jul 2009 02:38:58 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.998
X-Spam-Level: 
X-Spam-Status: No, score=-1.998 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, HTML_MESSAGE=0.001, J_CHICKENPOX_63=0.6]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id Kfg4ksGOLO+K for <tsvwg@core3.amsl.com>; Wed, 29 Jul 2009 02:38:57 -0700 (PDT)
Received: from wf-out-1314.google.com (wf-out-1314.google.com [209.85.200.172]) by core3.amsl.com (Postfix) with ESMTP id 350C53A6EE7 for <tsvwg@ietf.org>; Wed, 29 Jul 2009 02:38:49 -0700 (PDT)
Received: by wf-out-1314.google.com with SMTP id 28so185823wff.31 for <tsvwg@ietf.org>; Wed, 29 Jul 2009 02:38:49 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:date:message-id:subject :from:to:content-type; bh=Ryhf4paTRkyCrS/h6XqkNOziMgYqGn7JZjZjiqp+AOk=; b=EHBT71TI3xHuU9V6tICKdw7AsNNxKyAK9T81SZM62Y0T10fpCX50Wy7aw9s3FFgxDs DP8E55hMcva3So+xhAu/OaFiKD6Vxj5TX368wU7yxvwClNLsafYmH3xlWZWRJtaVi7zj QzPoQARERAPSmTRwyLNCAocuYhFZiSUdGVGks=
DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type; b=RRllUXgRJBwFo41vgSfmHrK/AQ3Lq5OVMpAb+h8yu9gYIETWhuT0OQ7M6wvqighqhv INM3AFkCTNVpN5Yirju6EpgTtUUioCmfum+MjdaFYEwqPheqlHv9ZrjqvOVQ8krXXJ7E ckCOZIWziqD09FWqooUUVV1pqE8cNYCOeClmA=
MIME-Version: 1.0
Received: by 10.143.9.21 with SMTP id m21mr912488wfi.172.1248860329641; Wed,  29 Jul 2009 02:38:49 -0700 (PDT)
Date: Wed, 29 Jul 2009 17:38:49 +0800
Message-ID: <d9eefe750907290238w587bc739mc5d069e2d17eeb49@mail.gmail.com>
From: Mohammad Arif <mohammad.arifb@gmail.com>
To: tsvwg@ietf.org
Content-Type: multipart/alternative; boundary=001636e0a46352875d046fd4f46f
Subject: [tsvwg] How to stop an sctp sender with zero A_rwnd
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 29 Jul 2009 09:38:58 -0000

--001636e0a46352875d046fd4f46f
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

Hi All
I m running my SCTP client with linux kernel sctp server.
I m sending SACK with A_rwnd size as zero for all DATA chunks,but the sender
still keeps on sending me DATA chunks.
As i have send A_Rwnd as zero, the peer  should stop sending me DATA chunks
?
Is my understanding wrong?

Please clarify
Thanks


-- 
Regards
Mohammad Arif

--001636e0a46352875d046fd4f46f
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div>Hi All</div>
<div>I m running my SCTP client with linux kernel sctp server.</div>
<div>I m sending SACK with A_rwnd size as zero for all DATA chunks,but the =
sender still keeps on sending me DATA chunks.</div>
<div>As i have send A_Rwnd as zero,=A0the peer =A0should stop sending me DA=
TA chunks ?</div>
<div>Is my understanding wrong?</div>
<div>=A0</div>
<div>Please clarify</div>
<div>Thanks</div>
<div><br clear=3D"all"><br>-- <br>Regards<br>Mohammad Arif </div>

--001636e0a46352875d046fd4f46f--

From bidulock@openss7.org  Wed Jul 29 03:00:05 2009
Return-Path: <bidulock@openss7.org>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 40AB73A6F28 for <tsvwg@core3.amsl.com>; Wed, 29 Jul 2009 03:00:05 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.999
X-Spam-Level: 
X-Spam-Status: No, score=-1.999 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, J_CHICKENPOX_63=0.6]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id B6GZvXe9RA6A for <tsvwg@core3.amsl.com>; Wed, 29 Jul 2009 03:00:04 -0700 (PDT)
Received: from gw.openss7.com (gw.openss7.com [206.75.119.236]) by core3.amsl.com (Postfix) with ESMTP id E5FF03A6F1D for <tsvwg@ietf.org>; Wed, 29 Jul 2009 03:00:03 -0700 (PDT)
Received: from wilbur.pigworks.openss7.net (IDENT:zi6pnGWhQhPhm3xhn2pbKrF0DWaLB9t5@ns5.evil.openss7.net [192.168.9.5]) by gw.openss7.com (8.13.8/8.13.8/Debian-3) with ESMTP id n6TA01I0006421; Wed, 29 Jul 2009 04:00:01 -0600
Received: from wilbur.pigworks.openss7.net (IDENT:AdnYzXIwXy35pHZlc09BibutTfGQWu4d@localhost [127.0.0.1]) by wilbur.pigworks.openss7.net (8.13.8/8.13.8/Debian-3) with ESMTP id n6TA01MS015076; Wed, 29 Jul 2009 04:00:01 -0600
Received: (from brian@localhost) by wilbur.pigworks.openss7.net (8.13.8/8.13.8/Submit) id n6TA0050015075; Wed, 29 Jul 2009 04:00:00 -0600
Date: Wed, 29 Jul 2009 04:00:00 -0600
From: "Brian F. G. Bidulock" <bidulock@openss7.org>
To: Mohammad Arif <mohammad.arifb@gmail.com>
Message-ID: <20090729100000.GA13094@openss7.org>
Mail-Followup-To: Mohammad Arif <mohammad.arifb@gmail.com>, tsvwg@ietf.org
References: <d9eefe750907290238w587bc739mc5d069e2d17eeb49@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <d9eefe750907290238w587bc739mc5d069e2d17eeb49@mail.gmail.com>
Organization: http://www.openss7.org/
Dsn-Notification-To: <bidulock@openss7.org>
X-Spam-To: <blockme@openss7.com>
User-Agent: Mutt/1.5.13 (2006-08-11)
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] How to stop an sctp sender with zero A_rwnd
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
Reply-To: bidulock@openss7.org
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 29 Jul 2009 10:00:05 -0000

Mohammad,

The SCTP sender is always permitted to have packet in flight.  The
data sent to you when you send arwnd=0 is probing the zero window
to detect when you have processed the enough messages to reopen the
window.

--brian

Mohammad Arif wrote:                       (Wed, 29 Jul 2009 17:38:49)
> 
>    Hi All
> 
>    I m running my SCTP client with linux kernel sctp server.
> 
>    I m sending SACK with A_rwnd size as zero for all DATA chunks,but the
>    sender still keeps on sending me DATA chunks.
> 
>    As i have send A_Rwnd as zero, the peer  should stop sending me DATA
>    chunks ?
> 
>    Is my understanding wrong?
> 
> 
> 
>    Please clarify
> 
>    Thanks
> 
>    --
>    Regards
>    Mohammad Arif

-- 
Brian F. G. Bidulock
bidulock@openss7.org
http://www.openss7.org/

From randall@lakerest.net  Wed Jul 29 05:01:51 2009
Return-Path: <randall@lakerest.net>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 849323A702E for <tsvwg@core3.amsl.com>; Wed, 29 Jul 2009 05:01:51 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.999
X-Spam-Level: 
X-Spam-Status: No, score=-1.999 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, J_CHICKENPOX_63=0.6]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id JMyH2vKMpTlY for <tsvwg@core3.amsl.com>; Wed, 29 Jul 2009 05:01:50 -0700 (PDT)
Received: from lakerest.net (unknown [IPv6:2001:240:585:2:213:d4ff:fef3:2d8d]) by core3.amsl.com (Postfix) with ESMTP id B549F3A702D for <tsvwg@ietf.org>; Wed, 29 Jul 2009 05:01:49 -0700 (PDT)
Received: from [10.1.1.53] ([10.1.1.53]) (authenticated bits=0) by lakerest.net (8.14.3/8.14.3) with ESMTP id n6TC1imN086830 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=NOT); Wed, 29 Jul 2009 08:01:44 -0400 (EDT) (envelope-from randall@lakerest.net)
DKIM-Signature: a=rsa-sha1; c=simple/simple; d=lakerest.net; s=mail; t=1248868905; h=Cc:Message-Id:From:To:In-Reply-To:Content-Type: Content-Transfer-Encoding:Mime-Version:Subject:Date:References: X-Mailer; b=nJSetXSEkpLxa7j2ATi1WuoEyrdmdUhQ/slNNpecb7vKkqfIhONEx0O fTmVHEA2gKeziaq6nfCUXSatxgyR5kA==
Message-Id: <CF657852-00B7-4DA2-B93D-D6E9D3A6D947@lakerest.net>
From: Randy Stewart <randall@lakerest.net>
To: Mohammad Arif <mohammad.arifb@gmail.com>
In-Reply-To: <d9eefe750907290238w587bc739mc5d069e2d17eeb49@mail.gmail.com>
Content-Type: text/plain; charset=US-ASCII; format=flowed; delsp=yes
Content-Transfer-Encoding: 7bit
Mime-Version: 1.0 (Apple Message framework v935.3)
Date: Wed, 29 Jul 2009 08:01:44 -0400
References: <d9eefe750907290238w587bc739mc5d069e2d17eeb49@mail.gmail.com>
X-Mailer: Apple Mail (2.935.3)
Cc: tsvwg@ietf.org
Subject: Re: [tsvwg] How to stop an sctp sender with zero A_rwnd
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 29 Jul 2009 12:01:51 -0000

Mohammad:

Yes and no..

It is true that a sender should stop sending you data... but not
all data.

Basically it should go in to a very slow sending/resending of data.
Since it must probe for when the rwnd is increased again.

The spec says formally:

"
A) At any given time, the data sender MUST NOT transmit new data to  
any destination transport address if its peer's rwnd indicates that  
the peer has no buffer space (i.e., rwnd is 0; see Section 6.2.1).  
However, regardless of the value of rwnd (including if it is 0), the  
data sender can always have one DATA chunk in flight to the receiver  
if allowed by cwnd (see rule B, below). This rule allows the sender to  
probe for a change in rwnd that the sender missed due to the SACK's  
having been lost in transit from the data receiver to the data sender.
"

Note the rule that says it can have 1 data chunk in flight (including  
if the rwnd is 0).

So its true bulk transfer should stop.. but you should still get
data.. and if you send back a sack acking that chunk you will
get the next TSN.. if you send back a sack acking the previous
chunk holding your rwnd at 0, then the TSN should be resent to
you

Hope that helps

R


On Jul 29, 2009, at 5:38 AM, Mohammad Arif wrote:

> Hi All
> I m running my SCTP client with linux kernel sctp server.
> I m sending SACK with A_rwnd size as zero for all DATA chunks,but  
> the sender still keeps on sending me DATA chunks.
> As i have send A_Rwnd as zero, the peer  should stop sending me DATA  
> chunks ?
> Is my understanding wrong?
>
> Please clarify
> Thanks
>
>
> -- 
> Regards
> Mohammad Arif

-----
Randall Stewart
randall@lakerest.net





From rbriscoe@jungle.bt.co.uk  Thu Jul 30 08:50:18 2009
Return-Path: <rbriscoe@jungle.bt.co.uk>
X-Original-To: tsvwg@core3.amsl.com
Delivered-To: tsvwg@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id D4C0A28C10C for <tsvwg@core3.amsl.com>; Thu, 30 Jul 2009 08:50:18 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.054
X-Spam-Level: 
X-Spam-Status: No, score=-3.054 tagged_above=-999 required=5 tests=[AWL=1.063,  BAYES_00=-2.599, DNS_FROM_RFC_BOGUSMX=1.482, GB_I_INVITATION=-2, RCVD_IN_DNSWL_LOW=-1]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 6EaJ1C+beD0r for <tsvwg@core3.amsl.com>; Thu, 30 Jul 2009 08:50:12 -0700 (PDT)
Received: from smtp1.smtp.bt.com (smtp1.smtp.bt.com [217.32.164.137]) by core3.amsl.com (Postfix) with ESMTP id BEFFE3A6C43 for <tsvwg@ietf.org>; Thu, 30 Jul 2009 08:50:08 -0700 (PDT)
Received: from i2kc08-ukbr.domain1.systemhost.net ([193.113.197.71]) by smtp1.smtp.bt.com with Microsoft SMTPSVC(6.0.3790.3959);  Thu, 30 Jul 2009 16:50:09 +0100
Received: from cbibipnt08.iuser.iroot.adidom.com ([147.149.100.81]) by i2kc08-ukbr.domain1.systemhost.net with Microsoft SMTPSVC(6.0.3790.3959); Thu, 30 Jul 2009 16:50:08 +0100
Received: From bagheera.jungle.bt.co.uk ([132.146.168.158]) by cbibipnt08.iuser.iroot.adidom.com (WebShield SMTP v4.5 MR1a P0803.399); id 1248969007554; Thu, 30 Jul 2009 16:50:07 +0100
Received: from mut.jungle.bt.co.uk ([10.73.65.160]) by bagheera.jungle.bt.co.uk (8.13.5/8.12.8) with ESMTP id n6UFnsuS023154; Thu, 30 Jul 2009 16:49:55 +0100
Message-Id: <200907301549.n6UFnsuS023154@bagheera.jungle.bt.co.uk>
X-Mailer: QUALCOMM Windows Eudora Version 7.1.0.9
Date: Thu, 30 Jul 2009 16:49:56 +0100
To: "BLACK, David" <Black_David@emc.com>
From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"; format=flowed
Content-Transfer-Encoding: quoted-printable
X-Scanned-By: MIMEDefang 2.56 on 132.146.168.158
X-OriginalArrivalTime: 30 Jul 2009 15:50:08.0421 (UTC) FILETIME=[6A117550:01CA112D]
Cc: tsvwg IETF list <tsvwg@ietf.org>
Subject: [tsvwg] Fwd: RE: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
X-BeenThere: tsvwg@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Transport Area Working Group <tsvwg.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/tsvwg>
List-Post: <mailto:tsvwg@ietf.org>
List-Help: <mailto:tsvwg-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/tsvwg>, <mailto:tsvwg-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 30 Jul 2009 15:50:18 -0000

David,

Can we try to resolve your concern over dropping=20
ECT(1) outer + Not-ECT inner, fairly soon?

I don't want to send a draft to IPsecME for=20
review until we're stable on at least the basic technical proposals.


Bob


>Date: Mon, 27 Jul 2009 00:35:59 +0100
>To: <Black_David@emc.com>
>From: Bob Briscoe <rbriscoe@jungle.bt.co.uk>
>Subject: RE: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
>Cc: <tsvwg@ietf.org>
>Bcc: =83\In 2009
>
>David,
>
>offlist
>
>At 14:28 26/07/2009, Black_David@emc.com wrote:
>>Bob,
>>
>>I think you may have gone back into the PCN-like behavior swamp
>>by proposing to distinguish ECT(0) from ECT(1).
>>
>>Alternatively, are you arguing that this distinction is
>>necessary in tandem with change b):
>>   [ECT(0) inner + ECT(1) outer -> ECT(1) in resulting header]
>>in order to make 3 ECN codepoints really usable?
>
>Of course. All the MUSTs in any draft should be=20
>read as a set of mutually supporting changes.
>
>
>
>
>Bob
>
>
>>Thanks,
>>--David
>>
>> > -----Original Message-----
>> > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
>> > Sent: Friday, July 24, 2009 4:49 AM
>> > To: Black, David
>> > Cc: tsvwg@ietf.org
>> > Subject: RE: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
>> >
>> > David,
>> >
>> > Thanks for the insight into the likely reasoning.
>> >
>> >  From a transport perspective, I think now is the right time to say
>> > safety needs greater attention...
>> > - there's a desire to change 4301 decap anyway to support 2
>> > severity levels,
>> > - safety's become more important since 4301 - greater chance
>> > of errors...
>> >
>> > Guys I've talked to in the pseudowire area are considering using ECN
>> > for middlebox to middlebox congestion control (a bit like PCN).
>> > Pseudowires (or PCN) might or might not also require IPsec. But it's
>> > quite an art to turn on ECN across a tunnelled segment and do no harm
>> > in both the following cases:
>> > - if e2e ECN is on, avoiding polluting e2e ECN signals
>> > - if e2e ECN is off, avoiding turning it on
>> > (see recent PCN list).
>> >
>> > This all means the chances that an error will occur increases, either
>> > in broken implementations or broken deployments.
>> >
>> > The new rules in ecn-tunnel are designed so that tunnels should "just
>> > work" and "safely", even if they are deployed by someone who doesn't
>> > know they have to cater for an in-path internal ECN loop as well.
>> >
>> > Q. What do you think of extending drop to both CE and ECT(1) outer if
>> > there's a Not-ECT inner (see table)?
>> > Reasoning: As these are both currently error conditions and likely to
>> > remain so, rather than two different safety rules for different
>> > circumstances just make tunnels safe in all potential cases.
>> > Otherwise someone has to remember to configure a different safety
>> > rule on all their tunnels when PCN or pseudowires get deployed in
>> > their network.
>> >
>> > +----------------+----------------------------------------------+
>> > | Incoming Inner |             Incoming Outer Header            |
>> > |     Header     +---------+------------+------------+----------+
>> > |                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
>> > +----------------+---------+------------+------------+----------+
>> > |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    | drop
>> >   |<-3168
>> > |    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |Not-ECT
>> >   |<-4301
>> > |    Not-ECT     | Not-ECT | Not-ECT    | drop       | drop
>> >   |<-ecn-tun
>> > +----------------+---------+------------+------------+----------+
>> >                   |                Outgoing Header               |
>> >                   +----------------------------------------------+
>> >
>> >
>> > Bob
>> >
>> > At 22:44 23/07/2009, Black_David@emc.com wrote:
>> > >Bob,
>> > >
>> > > > RFC4301 seems to say: If the inner is Not-ECT, anything
>> > else in the
>> > > > outer must be an error. But whereas RFC3168 dropped the packet if
>> > > > there was CE in the outer, RFC4301 says forward it as Not-ECT
>> > > > in all cases.
>> > > >
>> > > > RFC4301 gives no justification for this rather important
>> > change from
>> > > > RFC3168. I've spent the last hour or so trawling the
>> > IPsec and tsvwg
>> > > > archives and can't find any discussion on this point
>> > either. Was it
>> > > > discussed anywhere?
>> > >
>> > >Not that I recall.  I believe you're looking at a judgment call for
>> > >error case handling.  RFC 4301 was specified with close to the
>> > >minimum necessary to get ECN to work across an IPsec tunnel, as
>> > >security is an area where the default should be leaving out things
>> > >that aren't really necessary.  I believe that handling of this error
>> > >case was essentially treated as a "don't care."  Looking at your
>> > >alternative lines of thought:
>> > >
>> > > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
>> > > > then subsequently a congested router marked it CE, then
>> > it's unsafe
>> > > > to remove CE and forward Not-ECT - drop would have been safer.
>> > >
>> > >The ingress is rather unlikely to have put ECT in the outer - that
>> > >clearly violates both RFC 3168 and RFC 4301, and is a bad idea.
>> > >Hence something's almost certainly broken ...
>> > >
>> > > > * If something broken set CE on the outer even tho there was no
>> > > > congestion, it's safe.
>> > >
>> > >If the packet shows up with Not-ECT in the inner and CE in the
>> > >outer, something's definitely broken, even if the breakage is
>> > >actually a bad Not-ECT -> ECT transition followed by a correct
>> > >downstream ECT->CE transition.  Not attempting to perform a
>> > >compensating fix is a reasonable course of action, IMHO ...
>> > >but I could also defend the 3168 drop behavior as within
>> > >reason, along the lines that you suggest.
>> > >
>> > >Thanks,
>> > >--David
>> > >
>> > >
>> > > > -----Original Message-----
>> > > > From: Bob Briscoe [mailto:rbriscoe@jungle.bt.co.uk]
>> > > > Sent: Monday, July 20, 2009 7:48 AM
>> > > > To: Black, David
>> > > > Cc: tsvwg@ietf.org
>> > > > Subject: draft-ietf-tsvwg-ecn-tunnel: no drop on 4301 decap
>> > > >
>> > > > David,
>> > > >
>> > > > I wanted to move on from optional alarms to my proposed functional
>> > > > changes, ie. your case a).
>> > > >
>> > > > But I have to admit a big error which makes that moot. In
>> > > > ecn-tunnel-02 I incorrectly reported what RFC3168 &
>> > RFC4301 said (I
>> > > > just re-read them). See the table below.
>> > > >
>> > > > IP in IP Decapsulation; Recap of Pre-existing Behaviour
>> > > >+----------------+----------------------------------------------+
>> > > >| Incoming Inner |             Incoming Outer Header            |
>> > > >|     Header     +---------+------------+------------+----------+
>> > > >|                | Not-ECT |   ECT(0)   |   ECT(1)   |    CE    |
>> > > >+----------------+---------+------------+------------+----------+
>> > > >|    Not-ECT     | Not-ECT | drop       | drop       |
>> > drop     |<my error
>> > > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT    |
>> > drop     |<-3168
>> > > >|    Not-ECT     | Not-ECT | Not-ECT    | Not-ECT
>> > |Not-ECT   |<-4301
>> > > >|     ECT(0)     |  ECT(0) | ECT(0)     | ECT(0)     |   CE     |
>> > > >|     ECT(1)     |  ECT(1) | ECT(1)     | ECT(1)     |   CE     |
>> > > >|       CE       |      CE |     CE     |     CE     |   CE     |
>> > > >+----------------+---------+------------+------------+----------+
>> > > >                 |                Outgoing Header               |
>> > > >                 +----------------------------------------------+
>> > > >
>> > > > RFC4301 seems to say: If the inner is Not-ECT, anything
>> > else in the
>> > > > outer must be an error. But whereas RFC3168 dropped the packet if
>> > > > there was CE in the outer, RFC4301 says forward it as Not-ECT
>> > > > in all cases.
>> > > >
>> > > > RFC4301 gives no justification for this rather important
>> > change from
>> > > > RFC3168. I've spent the last hour or so trawling the
>> > IPsec and tsvwg
>> > > > archives and can't find any discussion on this point
>> > either. Was it
>> > > > discussed anywhere?
>> > > >
>> > > > This RFC4301 behaviour seems unsafe, but perhaps
>> > understandable. It
>> > > > all depends how the situation arose:
>> > > > * If the ingress wrongly turned a Not-ECT inner into an ECT outer,
>> > > > then subsequently a congested router marked it CE, then
>> > it's unsafe
>> > > > to remove CE and forward Not-ECT - drop would have been safer.
>> > > > * If something broken set CE on the outer even tho there was no
>> > > > congestion, it's safe.
>> > > >
>> > > >
>> > > >
>> > > > Bob
>> > > >
>> > > > At 12:07 13/07/2009, Bob Briscoe wrote:
>> > > > >David,
>> > > > >
>> > > > >We're getting somewhere (if we set aside (a) for now).
>> > > > >
>> > > > > > At 17:09 10/07/2009, Black_David@emc.com wrote:
>> > > > >> > >Quick summary of the four changes proposed in the draft:
>> > > > >> > >
>> > > > >> > >a) Inner: Not-ECT, Outer: ECT(0)
>> > > > >> > >         RFC 3168/4301: drop(!!!), Change: Not-ECT(!!!)
>> > > > >> > >b) Inner: ECT(0), Outer: ECT(1)
>> > > > >> > >         RFC 3168/4301: ECT(0), Change: ECT(1)
>> > > > >> > >c) Inner: ECT(1), Outer: ECT(0)
>> > > > >> > >         RFC 3168/4301: ECT(1), Change: ECT(1)(!!!)
>> > > > >> > >d) Inner: CE, Outer: ECT(1)
>> > > > >> > >         RFC 3168/4301: CE, Change: CE(!!!)
>> > > > >> > >
>> > > > >> > >If we were designing from scratch, all of these
>> > changes would
>> > > > >> > >be fine things to do.  We're not, and especially after the
>> > > > >> > >discussion of how this impacts RFC 2003 gets added, this
>> > > > >> > >draft will be proposing changes to some of the most widely
>> > > > >> > >deployed mechanisms in the Internet.  To me, that strongly
>> > > > >> > >suggests a conservative design approach - focus on
>> > what really
>> > > > >> > >needs to be changed, and resist them temptation to fix other
>> > > > >> > >things that would be "nice to have".  In other words, IMHO
>> > > > >> > >the appropriate criteria for "fixing it" is a very strong
>> > > > >> > >cast that "it's broken".
>> > > > >> > >
>> > > > >> > >This is an area of principled disagreement on goals between
>> > > > >> > >Bob and myself that list discussion is unlikely to resolve.
>> > > > >> > >My view of the above four changes is that only b) makes it
>> > > > >> > >over the "really needs to be changed" bar (and even then,
>> > > > >> > >only with a *much* stronger statement of the PCN-based
>> > > > >> > >rationale than is in the current draft, and Bob concurs
>> > > > >> > >with strengthening the rationale).
>> > > > >> > >
>> > > > >> > >Turning to the other three changes, a) is clearly described
>> > > > >> > >as "nice to have", enough said.
>> > > > >> > >
>> > > > >> > >For c) and d), I don't believe the optional alarms are
>> > > > >> > >justified based on ECN (RFC 3168) alone.  A key observation
>> > > > >> > >is that this draft does not make the < b) ECT(0) [Inner]
>> > > > >> > >to ECT(1) [Outer], forward with ECT(1) > case subject to an
>> > > > >> > >optional alarm.  What this means is that this draft not only
>> > > > >> > >enables three codepoints (ECT(0), ECT(1) and CE), but it
>> > > > >> > >*assumes* that they are ordered in terms of priority,
>> > > > >> > >severity, or something analogous.  That is unwarranted as
>> > > > >> > >a general design assumption, IMHO.
>> > > > >> > >
>> > > > >> > >For this, reason, I think discussion of which
>> > transitions are
>> > > > >> > >optionally alarm-able in this fashion belongs in
>> > the context of
>> > > > >> > >a protocol that actually orders these codepoints, not in a
>> > > > >> > >general document that would serve to "just constrain future
>> > > > >> > >standards," borrowing Toby's words ;-).
>> > > > >> > >
>> > > > >> > >I could see discussion of optional alarms in an
>> > Appendix that
>> > > > >> > >discusses 3-valued PCN as an example, but not in
>> > the main body
>> > > > >> > >of the draft.  In the main body, it would be fine
>> > to say that
>> > > > >> > >specific uses of the ECN codepoints may assign meanings that
>> > > > >> > >cause some of these transitions to be appropriate for alarms
>> > > > >> > >(and point to the Appendix for a PCN example), but it would
>> > > > >> > >not be ok to constrain the assignment of those meanings by
>> > > > >> > >making the ECT(0) [Inner] to ECT(1) [Outer] change not
>> > > > >> > >optionally alarm-able.
>> > > > >> > >
>> > > > >> > >Thanks,
>> > > > >> > >--David
>> > > > >> > >
>> > > > >> > >
>> > > > >> > > > -----Original Message-----
>> > > > >> > > > From: tsvwg-bounces@ietf.org
>> > [mailto:tsvwg-bounces@ietf.org]
>> > > > >> > > > On Behalf Of Bob Briscoe
>> > > > >> > > > Sent: Thursday, July 09, 2009 3:08 PM
>> > > > >> > > > To: Black, David
>> > > > >> > > > Cc: gorry@erg.abdn.ac.uk; tsvwg IETF list
>> > > > >> > > > Subject: Re: [tsvwg] review of
>> > > > >> > draft-ietf-tsvwg-ecn-tunnel-02 pt 1/2
>> > > > >> > > >
>> > > > >> > > > David,
>> > > > >> > > >
>> > > > >> > > > Bringing to the list as agreed. I'll reply in 2
>> > > > parts. Part 1/2...
>> > > > >> > > >
>> > > > >> > > > Your comments get to the nub of the proposed egress
>> > > > >> > protocol changes.
>> > > > >> > > > Great... inline...
>> > > > >> > > >
>> > > > >> > > > At 00:43 09/07/2009, Black_David@emc.com wrote:
>> > > > >> > > > >Bob,
>> > > > >> > > > >
>> > > > >> > > > >Sorry for the delay on this: things got entirely too
>> > > > >> > busy around the
>> > > > >> > > > >end of the quarter here.  Having re-read the draft from
>> > > > >> > scratch, I
>> > > > >> > > > >have a number of things to comment on ... I will be
>> > > > in Stockholm,
>> > > > >> > > > >and some of these items are going to need
>> > further discussion
>> > > > >> > > > >(private and/or public) in Stockholm.
>> > > > >> > > > >
>> > > > >> > > > >Starting with this table in Figure 4:
>> > > > >> > > > >
>> > > > >> > > >
>> > > > >> >
>> > > >
>> > >+------------------+----------------------------------------------+
>> > > > >> > > > >|  Incoming Inner  |             Incoming Outer Header
>> > > > >> >          |
>> > > > >> > > > >|      Header
>> > > > >> > +---------+------------+------------+----------+
>> > > > >> > > > >|                  | Not-ECT |   ECT(0)   |   ECT(1)   |
>> > > > >> >    CE    |
>> > > > >> > > >
>> > > > >> >
>> > > >
>> > >+------------------+---------+------------+------------+----------+
>> > > > >> > > > >|     Not-ECT      | Not-ECT |Not-ECT(!!!)|   drop(!!!)|
>> > > > >> > drop(!!!)|
>> > > > >> > > > >|      ECT(0)      |  ECT(0) | ECT(0)     | ECT(1)     |
>> > > > >> >   CE     |
>> > > > >> > > > >|      ECT(1)      |  ECT(1) | ECT(1)(!!!)| ECT(1)     |
>> > > > >> >   CE     |
>> > > > >> > > > >|        CE        |      CE |     CE     |     CE(!!!)|
>> > > > >> >   CE     |
>> > > > >> > > >
>> > > > >> >
>> > > >
>> > >+------------------+---------+------------+------------+----------+
>> > > > >> > > > >                    |                Outgoing Header
>> > > > >> >           |
>> > > > >> > > > >
>> > > > >> > +----------------------------------------------+
>> > > > >> > > > >
>> > > > >> > > > >My initial focus is on what needs to change from RFC
>> > > > >> > 3168/RFC 4301
>> > > > >> > > > >behavior and why.  Given that these two RFCs line up, I
>> > > > >> > think changes
>> > > > >> > > > >need strong motivation.  So, here are the four
>> > changes, and
>> > > > >> > > > my thoughts:
>> > > > >> > > > >
>> > > > >> > > > >a) Inner: Not-ECT, Outer: ECT(0)
>> > > > >> > > > >         3168/4301: drop (!!!)
>> > > > >> > > > >         Change: Not-ECT (!!!)
>> > > > >> > > > >         Rationale: "allow future standards actions
>> > > > to use this
>> > > > >> > > > >                 combination"
>> > > > >> > > > >         My view: This ain't broke, why are you
>> > fixing it?
>> > > > >> > > >
>> > > > >> > > > It is broke in the sense tunnel egresses unnecessarily
>> > > > >> > drop harmless
>> > > > >> > > > packets merely because they carry a "currently
>> > unused" (CU)
>> > > > >> > > > combination of codepoints. This is similar to the problem
>> > > > >> > ECN faced
>> > > > >> > > > with firewalls dropping packets unless they were Not-ECT.
>> > > > >> > > >
>> > > > >> > > > (a) wasn't originally part of the changes I was
>> > > > >> > proposing. But Anil
>> > > > >> > > > Agawaal convinced me it is worth doing. Reasoning: If we
>> > > > >> > are making a
>> > > > >> > > > change to decap anyway, let's get it as right as
>> > > > possible. If we
>> > > > >> > > > decide not to make any other changes to decap, I wouldn't
>> > > > >> > fight for
>> > > > >> > > > this one alone.
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > > >b) Inner: ECT(0), Outer: ECT(1)
>> > > > >> > > > >         3168/4301: ECT(0)
>> > > > >> > > > >         Change: ECT(1)
>> > > > >> > > > >         Rationale: Treat ECT(1) as more severe
>> > congestion
>> > > > >> > > > than ECT(0).
>> > > > >> > > > >         My view: Longer discussion below.
>> > > > >> > > > >
>> > > > >> > > > >c) Inner: ECT(1), Outer: ECT(0)
>> > > > >> > > > >         3168/4301: ECT(1)
>> > > > >> > > > >         Change: ECT(1) (!!!)
>> > > > >> > > > >         Rationale: "no currently legal transition" can
>> > > > >> > > > produce this combo.
>> > > > >> > > > >         My view: Not a good idea, among other reasons
>> > > > >> > > > because it's at
>> > > > >> > > > >                 odds with your "future
>> > standards action"
>> > > > >> > > > explanation for
>> > > > >> > > > >                 the first change.  I'd remove
>> > this change
>> > > > >> > > > and instead deal
>> > > > >> > > > >                 with the possibility of
>> > raising an alarm in
>> > > > >> > > > a PCN-specific
>> > > > >> > > > >                 draft rather than declaring
>> > this to be an
>> > > > >> > > > alarm-able combo
>> > > > >> > > > >                 in general.
>> > > > >> > > > >
>> > > > >> > > > >d) Inner: CE, Outer: ECT(1)
>> > > > >> > > > >         3168/4301: CE
>> > > > >> > > > >         Change: CE (!!!)
>> > > > >> > > > >         Rationale & My view: <same as previous change>
>> > > > >> > > >
>> > > > >> > > > c) & d) aren't PCN specific.
>> > > > >> > > >
>> > > > >> > > > Reasoning for adding alarms (also said this eariler
>> > > > on list in
>> > > > >> > > > conversation with Gorry):
>> > > > >> > > > If security folks have concerns over possible attacks
>> > > > >> > from "currently
>> > > > >> > > > unused" combinations, I want to give them a way
>> > to manage the
>> > > > >> > > > perceived threat without just dropping packets. RFC3168
>> > > > >> > recommended
>> > > > >> > > > alarms for CU combinations. In these cases, I believe
>> > > > >> > RFC3168 simply
>> > > > >> > > > overlooked that these are also CU combinations.
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > > >So, I would remove a) as not defensible (no good reason
>> > > > >> > for change),
>> > > > >> > > > >and I would push the discussion of adding alarms for c)
>> > > > >> > and d) into
>> > > > >> > > > >a PCN-specific document.
>> > > > >> > > >
>> > > > >> > > > c) & d) are not to do with PCN.
>> > > > >> > > >
>> > > > >> > > > My motivation is: While making change (b) to enable
>> > > > PCN, lets do
>> > > > >> > > > (a,c,d) to make the IP layer as clean as possible, to
>> > > > >> > ensure easier
>> > > > >> > > > future evolvability.
>> > > > >> > > >
>> > > > >> > > > I don't see the issue. It says SHOULD log and MAY
>> > > > alarm. So the
>> > > > >> > > > operator doesn't have to do anything.
>> > > > >> > > >
>> > > > >> > > > >The c) alarm in particular is only
>> > > > >> > > > >meaningful if PCN or something like it is in use.
>> > > > >> > > >
>> > > > >> > > > No. The combination in c) always was CU, and it's still
>> > > > >> > CU with PCN.
>> > > > >> > > > None of the proposed changes are PCN-specific.
>> > > > >> > > >
>> > > > >> > > > Even (b) is intended to also be useful for any 3-level
>> > > > >> > ECN while not
>> > > > >> > > > precluding 2-level ECN and the nonce.
>> > > > >> > > >
>> > > > >> > > > NB:  3 level =3D 2 'congested' levels + 1 level not
>> > > > >> > congested, e.g. VCP
>> > > > >> > > > or PCN which were both mentioned in RFC4774. This doesn't
>> > > > >> > count the
>> > > > >> > > > not-ECN-capable state, which makes 4 states.
>> > > > >> > > >
>> > > > >> > > > >I don't view d)
>> > > > >> > > > >as deserving of an alarm in general, as even though
>> > > > something has
>> > > > >> > > > >happened, no damage has been done - this is in
>> > line with the
>> > > > >> > > > >"be liberal in what you accept" principle of
>> > > > protocol design.
>> > > > >> > > >
>> > > > >> > > > That's why I'm proposing an optional alarm. Optional
>> > > > alarms *are*
>> > > > >> > > > liberal, but they allow operators a way to address
>> > > > >> > security concerns
>> > > > >> > > > about 'known unknowns' [Rumsfelt], while still
>> > allowing new
>> > > > >> > > > behaviours to be introduced.
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > > >That leaves b), which I think is the key desired
>> > > > change, because
>> > > > >> > > > >it makes a third ECN code point usable through tunnels.
>> > > > >> > > >
>> > > > >> > > > Yes - exactly.
>> > > > >> > > >
>> > > > >> > > > >Much as
>> > > > >> > > > >this might be useful for PCN, it may be too late to
>> > > > do this.  The
>> > > > >> > > > >problem starts from the need to say that this draft
>> > > > >> > updates RFC 4301.
>> > > > >> > > > >The result will be a mix of 4301 and updated-4301
>> > > > implementations
>> > > > >> > > > >that render the third state difficult to use in
>> > > > practice.  The
>> > > > >> > > > >problem's even worse for non-IPsec tunnels.
>> > > > >> > > >
>> > > > >> > > > The idea is that an operator wanting to implement
>> > > > PCN will have a
>> > > > >> > > > spec they can point to in their invitation to
>> > > > tender. At the same
>> > > > >> > > > time this spec both UPDATEs and is backward compatible
>> > > > >> > with RFC3168 &
>> > > > >> > > > RFC4301 because it only changes CU stuff at egress and
>> > > > >> > the change at
>> > > > >> > > > the ingress leads to the same output at the egress.
>> > > > >> > Therefore vendors
>> > > > >> > > > can update 3168 & 4301 tunnels lazily - only when their
>> > > > >> > customers ask
>> > > > >> > > > for the new features (whether for PCN or for something
>> > > > >> > else like VCP).
>> > > > >> > > >
>> > > > >> > > > >I also read Appendix D
>> > > > >> > > > >as saying that PCN is going the two-state
>> > route, not three.
>> > > > >> > > >
>> > > > >> > > > That's not how it should be read - I'll clarify it. The
>> > > > >> > PCN w-g wants
>> > > > >> > > > 3 states, but can't have them while ecn-tunnel is
>> > > > not proposed
>> > > > >> > > > standard. Therefore, it has decided to define a 2-state
>> > > > >> > baseline and
>> > > > >> > > > a number of experimental schemes to work round this
>> > > > tunnel issue
>> > > > >> > > > (plus a comparison doc about them all). It has agreed
>> > > > >> > that eventually
>> > > > >> > > > just one experimental scheme will be chosen for
>> > > > standards track:
>> > > > >> > > > - 3-in-1 (the simplest, which depends on (b) of
>> > ecn-tunnel)
>> > > > >> > > > - 3-state (which uses an extra DSCP)
>> > > > >> > > > - psdm (which requires out of band state)
>> > > > >> > > > - sm (single marking gets 3 functions from 2
>> > > > encoding states but
>> > > > >> > > > admits admission control & termination aren't
>> > very precise in
>> > > > >> > > > all cases)
>> > > > >> > > >
>> > > > >> > > > ___
>> > > > >> > > > See part 2/2 for responses to your comments on
>> > specific parts
>> > > > >> > > > of the test.
>> > > > >> > > >
>> > > > >> > > > Thanks again for the review
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > > Bob
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > >
>> > > > ________________________________________________________________
>> > > > >> > > > Bob Briscoe,               Networks Research Centre,
>> > > > BT Research
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> > > >
>> > > > >> >
>> > > > >> >
>> > ________________________________________________________________
>> > > > >> > Bob Briscoe,               Networks Research Centre,
>> > BT Research
>> > > > >> >
>> > > > >> >
>> > > > >> >
>> > > > >
>> > > > >________________________________________________________________
>> > > > >Bob Briscoe,               Networks Research Centre, BT Research
>> > > >
>> > > > ________________________________________________________________
>> > > > Bob Briscoe,               Networks Research Centre, BT Research
>> > > >
>> > > >
>> > > >
>> >
>> > ________________________________________________________________
>> > Bob Briscoe,               Networks Research Centre, BT Research
>> >
>> >
>> >
>
>________________________________________________________________
>Bob Briscoe,               Networks Research Centre, BT Research

________________________________________________________________
Bob Briscoe,               Networks Research Centre, BT Research=20

