From dhcp-v6@bucknell.edu  Wed Dec  1 01:52:39 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id BAA11763
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 01:52:38 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id BAA07411;
	Wed, 1 Dec 1999 01:50:22 -0500 (EST)
Date: Wed, 1 Dec 1999 01:50:22 -0500 (EST)
Message-Id: <199912010604.BAA0000018396@quarry.zk3.dec.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Jim Bound <bound@zk3.dec.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DISCUSSION: Use of Address Extensions with DHCPv6 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

Charlie,

I don't agree with 90% of what you object to.  But for now I am just
going to listen to where the working group goes with this discussion.
especially the implementors.  Then you, Mike, and I can look at it and
state what we think we heard everyone say and present the issues.
So I am not going to defend my statements you don't agree with at this
point I would rather see the wg respond.

thanks
/jim


From dhcp-v6@bucknell.edu  Wed Dec  1 01:52:55 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id BAA11919
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 01:52:53 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id BAA30337;
	Wed, 1 Dec 1999 01:51:58 -0500 (EST)
Date: Wed, 1 Dec 1999 01:51:58 -0500 (EST)
Message-Id: <199912010645.BAA0000023661@quarry.zk3.dec.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Jim Bound <bound@zk3.dec.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

Hi Ralph,

>>This is idempotent and I can see no where we break the multiple server
>>model.  The protocol keeps on ticking whether there is 1...N links and
>>1...N servers.  

>It doesn't break the multiple server model, but (as I interpret the text)
>does not explicitly point out that the client may interact with multiple
>servers.  This might be a good place to add text about how the spec
>allows for concurrent interaction with multiple servers.

I think I may have missed where your going:

Are you going here:

client for le0 on 3ffe::12::12/48 (just one interface):

Client can send Requests to any server that it may have learned of or
about from the multicast solicit?  Thanks...

Nothing in the protocol prohibits this behavior but we don't call it out
either.

>>4.3. DHCP Request Message Format
>>
>>   In order to request configuration parameters from a server, a client
>>   sends a DHCP Request message, and MAY append extensions [15].  If
>>   the client does not know any server address, it MUST first obtain
>>   one by multicasting a DHCP Solicit message (see Section 4.1).
>>
>>Here is where we might add that this solicit in fact will support
>>multiple instants of solicit on different links (subnets in IPv4 terms)
>>or in section 4.1.  
>
>
>OK...

But this is different than my question above here it means:

client at le0 3ffe::12:12/48 sends a solicit on that link and hears
from severs.
client at le1 4ffe::12::12 (different link) sends a solicit on that link
and hears from servers.

Nothing in the protocol prohibits this behavior but we don't call it out
either.

>>5.3. Receiving DHCP Advertise Messages
>>
>>   After a client has received a DHCP Advertise message, it has the
>>   address of a server for subsequent DHCP Request messages.
>>
>>   [...]
>>
>>   Unless a DHCP Advertisement is received with a preference equal
>>   to 255 (see Section 4.2), the client MUST wait CLIENT_ADV_WAIT
>>   seconds after issuing the DHCP Solicit message in order to receive
>>   the Advertisement with the highest preference.  After waiting for
>>   that period of time, a client MUST select the highest preference
>>   server as the target of its DHCP request.  If a client receives an
>>   advertisement with a preference of 255, it does not have to wait for
>>   any more advertisements.
>>
>>Here what we are saying was an addition from the Chicago meeting.
>>
>>We assume the client on the same link could hear from multiple servers
>>as we were told by the working group in Chicago.  Per the working group
>>we added the preference field to the Advertisement message.  The way the
>>code and implementation could work is as follows:
>>
>>  boolean head_cheese_server = FALSE;
>>  int *head_cheese_server_data
>>  pthread_start_client_solicit;  /* this is not how it looks but */
>>  time_interrupt_set (CLIENT_ADV_WAIT);
>>  while (pthread_CLIENT_ADV_WAIT_not_back) {
>>    get_advertisement_message(server_msg_block);
>>    If (server_msg_block->preference_value = 255) {
>>       head_cheese_server = TRUE;
>>       head_cheese_server_data = server_msg_block;
>>       break;
>>    } else 
>>        store_this_server_for_later_sort (server_msg_block);
>>  }
>>  /* not putting in check here for discussion if no advertisements
>>     came back */
>>  If (head_cheese_server)
>>   use_the_head_cheese_server_for_request(server_msg_block);
>>  else
>>   sort_and_pick_highest_preference_server(server_msg_block);
>>
>>But the point is that we listen to multiple servers and select among
>>them based on the preference field from the server unless a server says
>>I am the boss or head cheese use me.  The first server to be heard that
>>does this wins.

>I don't see how this mechanism allows for use of more than one server.

Ok.  I am listening.

>The mechanism explicitly says "pick the one head_cheese or highest
>priority server".  What crieria does the client use to decide to use
>and choose a second server? 

Right now we don't specify that but we could?  If the client keeps the
server responses.  

Do you think we should define a mechanism to select a second, or even
third server?

Off the top of my head I would say I would use the preference ordinal
value?  

>>The client may store all server advs for later use too.
>>
>>   If a client sends a DHCP Request to a highly preferred server but
>>   fails to receive a DHCP reply from that server after following the
>>   retransmission algorithm in section 8, the client MUST then try to
>>   send a DHCP Request to a less preferred server.
>>
>>>(I think this section, at least, needs to be rewritten to allow interaction
>>>with multiple servers.  Under what conditions may a client disregard the
>>>server preference and begin a transaction with a second server with a
>>>[presumably] lower preference?)
>
>>So do you believe we should say the client SHOULD store servers that
>>respond to the solicit and then discuss the case when the client will
>>use a server of lower preference?
>>
>>This I believe resolves your concern and is doable?

>No, this doesn't exactly address my concern.  What I read is a mechanism
>for *serial contact* with multiple servers, but *interaction* with only
>one server at a time; i.e., the client works down the list, one server at a
>time, until it successfully completes a transaction with *one* server.

The protocol selects a server yes.  The criteria we have agreed to thus
far in the WG is to use the preference field to pick that server.
Once the client has done that selection for that interface it is the
server it uses.  For another interface the process begins again.  Each 
associated client interface on a link will spawn a new solicit.

Why would a client want to change servers if its working is what I am
wrestling with?

But if the client's server is not responding it can use a server learned
from the advertisements or from a new solicit.

Why would a client want to speak with one server for one set of
parameters and another server for another set of parameters?  Now we are
getting into SLP space I think.??

But maybe I am missing some data point?                         

thanks
/jim

- Ralph


From dhcp-v6@bucknell.edu  Wed Dec  1 07:43:10 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA06336
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 07:43:09 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA25858;
	Wed, 1 Dec 1999 07:41:33 -0500 (EST)
Date: Wed, 1 Dec 1999 07:41:33 -0500 (EST)
Message-Id: <Version.32.19991130195129.00e3c600@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 10:02 AM 11/30/99 -0500, you wrote:
>>Potentially open issues in the use of multiple servers:
>
>>* Text that implies interaction with a single server through use of
>>singular reference to "server" and through omission of any specific
>>reference to multiple servers
>
>This is pretty clear and pretty easy to fix IMO.

I agree.

>>* Semantics of Solicit message in the presence of multiple DHCP services:
>>how does a client know when it should interact with multiple servers; how
>>does a client differentiate between redundant servers (from the same
>>service) and servers with different resources?
>
>Just sent suggestion to fix that which will support getting this done
>rather expediently in the spec.  

Do you mean the prefix to be associated with a resource?  Or is the prefix
part of the message header?  If the latter is true, then the client will
be able to identify different groups of equivalent servers from which it
has received responses?

How does the client decide which group or groups of servers to contact?
I could imagine popping up a dialog box to the user - "You can connect to
these services - which do you want to connect to?" (where the "services"
are the groups of servers that have offered addresses; I tried to write
the dialog in English rather than talking about "offered addresses").

What is the mapping from a server prefix to some human-readable name?

- Ralph





From dhcp-v6@bucknell.edu  Wed Dec  1 07:43:16 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA06393
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 07:43:16 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA13101;
	Wed, 1 Dec 1999 07:41:41 -0500 (EST)
Date: Wed, 1 Dec 1999 07:41:41 -0500 (EST)
Message-Id: <Version.32.19991130195947.00e49b80@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 10:02 AM 11/30/99 -0500, Jim Bound wrote:
>>[...]it may
>>be that we need to clarify that a client needs to send multiple Solicit
>>messages, one to each potential DHCP service, to find the highest
>>preference server from each DHCP service.  
>
>The solicit is a multicast sendmsg() for a link (subnet) then the client
>will hear from multiple servers, which as you pointed out in other mail
>we need to clarify for the multiple server case.

OK, I guess we need to agree on an abstraction for a "DHCP service" and define
a mechanism for identifying services.  Then, we need to fix 5.3 to state
that a client applies the server selection algorithm to each service and
can contact one appropriate server from each service.

- Ralph




From dhcp-v6@bucknell.edu  Wed Dec  1 07:43:22 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA06447
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 07:43:21 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA22201;
	Wed, 1 Dec 1999 07:41:50 -0500 (EST)
Date: Wed, 1 Dec 1999 07:41:50 -0500 (EST)
Message-Id: <Version.32.19991130200215.00ded9f0@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 12:24 PM 11/30/99 -0500, Thomas Narten wrote:
>I think we are actually in agreement. My use of "does not support"
>appears to have been confusing and misleading. The underlying message
>formats certainly could support a client having multiple
>interactactions with different servers simultaneously.  However, the
>spec itself has little text on this and my expectation of what folks
>would actually implement (based on my read of what the current text
>says should be implemented) is that clients will in practice pick one
>server and interact only with it.

Ah, yes, we are in agreement.

- Ralph





From dhcp-v6@bucknell.edu  Wed Dec  1 07:43:33 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA06538
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 07:43:33 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA04842;
	Wed, 1 Dec 1999 07:42:01 -0500 (EST)
Date: Wed, 1 Dec 1999 07:42:01 -0500 (EST)
Message-Id: <Version.32.19991130200607.00e4f420@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 10:46 AM 11/30/99 -0800, you wrote:
>Probably it is better to describe states as pertaining to a
>transaction, than to a client.  Maybe it would be enough to
>specify that any client has to be able to run several state
>independent state machines simultaneously.  Maybe it would
>be even better to specify that state variables should not have
>global process scope.  Thread variables might have to be indexed
>into arrays per thread or something.

I don't think we want to go into those implementation details.
In fact, I don't think a client *has* to implement multiple
state machines for correctness.  I could imagine a single client
thread managing multiple interfaces with an event-driven
implementation.

>On the other hand, I think that many protocols can be specified
>in a way that is simple enough to allow correct implementations
>that don't rely on state machines.

In fact, we never intended to *mandate* a state machine for a
DHCPv4 implementation.  It seemed like a good way to describe the
client's behavior but not (necessarily) an implementation.

- Ralph





From dhcp-v6@bucknell.edu  Wed Dec  1 08:08:33 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA06337
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 07:43:09 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA02302;
	Wed, 1 Dec 1999 07:40:59 -0500 (EST)
Date: Wed, 1 Dec 1999 07:40:59 -0500 (EST)
Message-Id: <Version.32.19991130194715.00e3ece0@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 10:01 AM 11/30/99 -0500, Jim Bound wrote:
>Mike has proposed to add an address prefix of topological significance
>to the address extension.  The only realeasable resource we have right
>now.  For other reasons but close to the same reason.
>This could be added to also support knowing if a server is within the
>administrative domain as Barr suggested.

Jim, Mike: I can't find the message or messages with this proposal.  I
don't understand what you mean by "an address prefix of topological
significance".  And, would this prefix be part of the server's IPv6
address or a new field in the address extension or ???

>The way this would work is on the Advertisement we would add a field for
>the server to put its prefix topological significance.  When the client
>goes to release a resource to a server the address would be checked
>against that stored value for servers and provide the client knowledge
>if in fact that server can release the address.
>
>This is more doable than using a service bit mask for now and if we in
>the future define other than addresses we can use the server prefix as
>part of that releasable resources associations.  

I read this to explain that the prefix would identify groups of equivalent
servers, and that a client would use the prefix associated with a resource
to identify which of the servers it knows about can accept the released
resource.

How does the client learn about other servers to which it might release
a particular resource?

- Ralph





From dhcp-v6@bucknell.edu  Wed Dec  1 09:05:51 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA15822
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 09:05:50 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id JAA20904;
	Wed, 1 Dec 1999 09:01:02 -0500 (EST)
Date: Wed, 1 Dec 1999 09:01:02 -0500 (EST)
Message-Id: <199912011303.IAA01018@grosse.manhattan.fugue.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ted Lemon <mellon@isc.org>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.


> Would you be willing to support my contention that, instead, the
> state machine be associated with the transaction by which the
> IP address is expected to be obtained?  This seems to avoid some
> of the possible negative interaction between independent requests.

Transactions are ephemeral, so I don't see how you can associate state
machines with them.   Am I misunderstanding what you are saying here?
When I said that the xid is used to associate an incoming event with a
state machine, I didn't say that of course the xid changes with each
transaction, so once the state has changed, and incoming
(retransmitted) packet associated with a particular state machine
would no longer be recognized as being associated with that state
machine, and would have to be ignored.   I suppose that given that in
DHCPv6 the server can initiate a transaction, the transaction ID by
itself couldn't be used to associate a given transaction to a given
state machine.

			       _MelloN_


From dhcp-v6@bucknell.edu  Wed Dec  1 13:30:31 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA26728
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 13:30:29 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA01583;
	Wed, 1 Dec 1999 13:25:04 -0500 (EST)
Date: Wed, 1 Dec 1999 13:25:04 -0500 (EST)
Message-Id: <199912011802.NAA0000001871@quarry.zk3.dec.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Jim Bound <bound@zk3.dec.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Clarification
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

WG Members,

Need to provide some clarification and whats happening here.

> 
> Charlie,
> 
> I don't agree with 90% of what you object to.  But for now I am just
> going to listen to where the working group goes with this discussion.
> especially the implementors.  Then you, Mike, and I can look at it and
> state what we think we heard everyone say and present the issues.
> So I am not going to defend my statements you don't agree with at this
> point I would rather see the wg respond.
> 
> thanks
> /jim

Please don't take this that Charlie, Mike, and I are not working
together.  What has happened is we are getting really great input and
addressing (no pun intended) the issues Mike so clearly artitulated that
we must resolve to get DHCPv6 to PS.  In some cases its happening
realtime and the intial discussion is taking place on the list. My
disagreement with Charlie are fine in this space and we always converge
and we usually do this offline but we don't want to stop the momentum
eitiher.  I proposed an addressing model for clients to get addresses to
resolve the issues from Mikes summary and that we have changed the
dhcpv4 and addrconf model.  I propose a compromise and also that we do
get into very detailed description of what happens for each case in this
part of the spec.  In our zest to work on this we did not have time to
talk offline.  We are discussing this and other ideas now as authors and
will converge but please keep this part alive it is the most important
part of the change to dhcpv6 IMO that we get right and consensus on for
initial success to submit as PS.

Also if you have sent us offline input on the addressing discussion
please send it to the DHCPv6 list so all see it.

The authors are in synch and hard at work in addition to the great set
of mail exchange on all the topics.  Its like having two jobs here.

I am dropping balls at my day job and very attentive to this process
there is nothing more important going on for me than DHCPv6 for a little
while anyway.

thanks so much,
/jim


From dhcp-v6@bucknell.edu  Wed Dec  1 14:22:45 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA23487
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 14:22:44 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id OAA26558;
	Wed, 1 Dec 1999 14:20:05 -0500 (EST)
Date: Wed, 1 Dec 1999 14:20:05 -0500 (EST)
Message-Id: <199912011822.NAA05778@rotala.raleigh.ibm.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Thomas Narten <narten@raleigh.ibm.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

Charlie,

> Ralph Droms wrote:

> > Put succinctly, I know of no abstract reason to preclude the concurrent
> > interoperation of a DHCP client with multiple servers.  In my opinion, the
> > current spec does not adequately describe the details of concurrent use of
> > multiple servers to ensure interoperable DHCPv6 clients that will provide
> > predictable, manageable, useful behavior.

> I hope you be willing to point out text that is wrong, or suggest
> some strategy for us to forestall erroneous interpretations.

I have been trying (for quite some time) to say that there are parts
of the spec that are unclear or ambiguous. I've also pointed out
specific text. This goes as far back as my 9/98 review.  However, I
sense that you simply disagree and don't see the need for changes. If
my concerns about the overall clarity of the spec were just my own
personal opinion, I'd let it drop. But in speaking with others, I hear
similar concerns and concurrance that the issues I've raised are
valid. Hence, I can't just let them drop. I read your response to
Ralph's specific comments as just saying "no, the spec is fine, there
is no need for changes".  I don't have the cycles to keep circling the
bush on this one.  We are just going to have to agree to disagree on
this point.

Thomas


From dhcp-v6@bucknell.edu  Wed Dec  1 15:22:08 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA21037
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 15:22:04 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA30262;
	Wed, 1 Dec 1999 15:17:35 -0500 (EST)
Date: Wed, 1 Dec 1999 15:17:35 -0500 (EST)
Message-Id: <384579E3.CBA6F0AC@iprg.nokia.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Charles E. Perkins" <charliep@iprg.nokia.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
Content-Transfer-Encoding: 7bit


Hello Thomas,

> > I hope you be willing to point out text that is wrong, or suggest
> > some strategy for us to forestall erroneous interpretations.
> 
> I have been trying (for quite some time) to say that there are parts
> of the spec that are unclear or ambiguous. I've also pointed out
> specific text. This goes as far back as my 9/98 review.  However, I
> sense that you simply disagree and don't see the need for changes. If
> my concerns about the overall clarity of the spec were just my own
> personal opinion, I'd let it drop. But in speaking with others, I hear
> similar concerns and concurrance that the issues I've raised are
> valid. Hence, I can't just let them drop. I read your response to
> Ralph's specific comments as just saying "no, the spec is fine, there
> is no need for changes".  I don't have the cycles to keep circling the
> bush on this one.  We are just going to have to agree to disagree on
> this point.

Actually, I don't want to be disagreeable at all, and I do not
object to changing the specification to make it clearer.  However,
I don't know exactly what isn't clear, so I cannot myself suggest
the changes.  I reckon that Mike and Jim may have more of an idea
about why the current wording is unclear, and so I will look forward
to suggestions from them or anyone else about how to make the
existing wording clearer.  As long as the intended meaning does
not change, I have only encouragement for the effort.

Regards,
Charlie P.


From dhcp-v6@bucknell.edu  Wed Dec  1 17:31:51 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA24744
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 17:31:47 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id RAA32364;
	Wed, 1 Dec 1999 17:26:08 -0500 (EST)
Date: Wed, 1 Dec 1999 17:26:08 -0500 (EST)
Message-Id: <38459964.406AA211@iprg.nokia.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Charles E. Perkins" <charliep@iprg.nokia.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: Summary: DC meeting question #1
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
Content-Transfer-Encoding: 7bit


Hello Mike,

I hope you don't mind if I put spaces in different places in your
reply.  It helped me to read your note better, and to read my responses
to your note better too.

> 1) General Form
> 
>     *   Releasable resource extensions need to have a general form so that
>         clients/servers can manage them in a generic fashion.

Sounds good.

>     *   There is a generic bit in the extensions header which is set if the
>         extension is carrying a releasable resource. When this bit is set,
>         there are two 32 bit fields present in the extension which represent
>         the relative times in seconds when the client should start attempting
>         to extend the lease on the resource, and when the client must stop
>         using the resource if it is unable to extend it.

In Minneapolis, the suggestion was made that one range of extension
types was releasable, and that the rest of the extension types were
nonreleasable.  There seemed to be some general approval for this idea;
I don't remember who proposed it.

If the extensions from 32767-65535 were reserved for releasable
resources, that would work.  There could be a section in the
document prescribing a format for these extensions, but in general
I think that clients that ask for releasable resources can be
expected to know a lot about those resources.  Thus, working out
a general format for these future extensions may not be overwhelmingly
necessary right now.

>     *   Since extensions carrying releasable resources have a general
>         form/semantics, clients/servers do not need to change when new
>         extensions are defined which are releasable.

I don't totally disagree with this statement, but as I just
described it seems likely that consuming a releasable resource
implies some other systemic changes in the client that are more
fundamental than any standardized message format.

 are they allocated/requested, and how and to whom are they
> released?
> 
>     *   Servers advertise releasable resources. We (WG) need to consider
>         how releasable resources are advertised.

If a client really requires a particular releasable resource, we
should allow it to append an extension to its solicitation, with
the meaning that it wants an advertisement from a server that can
offer the resource.
 
>     *   Clients MUST request releasable resources, otherwise the server
>         will not allocate them.

This matches the general agreement I remember from Minneapolis.

>     *   Servers allocate releasable resources for a period of time set by
>         administrative policy

Check.

>     *   A client can renew releasable resources w/ allocating server, after
>         the first timer expires. If unable to contact the server, the client
>         must stop using resource when second timer expires.

Check, according to the proposed generalized scheme.  We should allow
for possible variations depending upon the particular resource that
is eventually designed for allocation by way of DHCP.  If a client
needs a "sophisticated" resource, we shouldn't proscribe such
sophisticated
clients from getting service from suitably sophisticated servers.

>     *   A client SHOULD release releasable resources to the allocating server.
>         If it fails to contact the allocating server, it MAY locate another DHCP
>         server and attempt to release them. In any case, the server will recover
>         the resource when its lease has expired. Thus, like DHCPv4, releasable
>         resources (addresses) will ultimately be reclaimed as administrative
>         policy dictates, and it isn't critical for clients to contact the server.

Check.

> 3) Multiple Server Environments
> 
>     *   Servers advertise available IP address(es) and other releasable
>         resources, thus clients know what resources are available from which
>         server.

Agreed, but in response to specific client requests only.  Otherwise,
the advertisements could get pretty long.  Case in point, what if a
server had to advertise every prefix from which it could allocate
addresses?  Or every disk farm from which it could allocate scratch
space?  Or ...?

>     *   We should avoid over specifying how DHCP works within multi-server
>         environments. However, an introductory/scenario section w/ a protocol
>         timeline would help validate the protocol model, and give some guidance
>         to implementors where needed.

Check.

> Miscellaneous:
> 
> *       I think we do need an "extension request extension" - having such an
>         extension would simplify the advertisement of informational extensions,
>         as well as simply a client's method for requesting extensions.
>         Note: We need to specify that a policy switch should be available
>               in the server to determine when an extension is returned (because
>               of administrative policy, or because client requested it?).
>         This point has been a bone of contention in DHCPv4, and we should
>         endeavor to clarify the semantics in DHCPv6.

Agreed on the suitability of such an extension for this purpose.

> *       Erik raised an interesting point about application layer extensions on
>         machines multiple interfaces: If you have multiple NTP servers 
>         available from DHCP information from different interfaces, which one
>         do you select? In our (Sun's) DHCPv4 client implementation, we have
>         a notion of a "primary" interface, from which application
>         layer parameters take precedence over those set in other interfaces.
>         It seems to me that a similar DHCPv6 implementation strategy could
>         resolve this issue. Charlie suggested that the server preference
>         value be used to determine preference. His suggestion has the
>         advantage that preference is set by the administrator. I prefer
>         the former solution ("primary" interface).

Perhaps for extensions where it actually matters, the preference could
be a field of the extension itself.  My original suggestion was just
on the spur of the moment, and not made with very much real study.
Letting the client decide is probably O.K. in most instances, I reckon,
so if it always decides in favor of a "primary interface", no harm is
likely to be done.

Regards,
Charlie P.


From dhcp-v6@bucknell.edu  Wed Dec  1 23:43:17 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA01903
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 23:43:15 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id XAA01349;
	Wed, 1 Dec 1999 23:40:06 -0500 (EST)
Date: Wed, 1 Dec 1999 23:40:06 -0500 (EST)
Message-Id: <199912020423.XAA0000008344@quarry.zk3.dec.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Jim Bound <bound@zk3.dec.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.


folks...I am catching up now and reading serially and responding...
got yanked this p.m. 

>At 10:01 AM 11/30/99 -0500, Jim Bound wrote:
>>Mike has proposed to add an address prefix of topological significance
>>to the address extension.  The only realeasable resource we have right
>>now.  For other reasons but close to the same reason.
>>This could be added to also support knowing if a server is within the
>>administrative domain as Barr suggested.
>
>Jim, Mike: I can't find the message or messages with this proposal.  I
>don't understand what you mean by "an address prefix of topological
>significance".  And, would this prefix be part of the server's IPv6
>address or a new field in the address extension or ???

[If mike has not already sent it].  This is an idea Mike sent around and
was to present at Wash D.C. but we did not get to that bullet.  I will
let Mike resend it was in one of his summaries on resolution.  Mike can
find it quicker than I me thinks :---)...

>>The way this would work is on the Advertisement we would add a field for
>>the server to put its prefix topological significance.  When the client
>>goes to release a resource to a server the address would be checked
>>against that stored value for servers and provide the client knowledge
>>if in fact that server can release the address.
>>
>>This is more doable than using a service bit mask for now and if we in
>>the future define other than addresses we can use the server prefix as
>>part of that releasable resources associations.  

>I read this to explain that the prefix would identify groups of equivalent
>servers, and that a client would use the prefix associated with a resource
>to identify which of the servers it knows about can accept the released
>resource.

Yep.

>How does the client learn about other servers to which it might release
>a particular resource?

I think thats the absraction we need from the previous mail thread I
just responded too.  We would have to have a means to identify this.

/jim  



From dhcp-v6@bucknell.edu  Wed Dec  1 23:43:19 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA01902
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 23:43:15 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id XAA03794;
	Wed, 1 Dec 1999 23:37:26 -0500 (EST)
Date: Wed, 1 Dec 1999 23:37:26 -0500 (EST)
Message-Id: <199912020413.XAA0000025921@quarry.zk3.dec.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Jim Bound <bound@zk3.dec.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.


>>>* Semantics of Solicit message in the presence of multiple DHCP services:
>>>how does a client know when it should interact with multiple servers; how
>>>does a client differentiate between redundant servers (from the same
>>>service) and servers with different resources?
>>
>>Just sent suggestion to fix that which will support getting this done
>>rather expediently in the spec.  
>
>Do you mean the prefix to be associated with a resource?  Or is the prefix
>part of the message header?  If the latter is true, then the client will
>be able to identify different groups of equivalent servers from which it
>has received responses?

I had mean't the former but now as I read you mail I think your right we
need to do the latter.  OK let me bounce this off Charlie and Mike. And
hope others comment too.  But good suggestion.

>How does the client decide which group or groups of servers to contact?
>I could imagine popping up a dialog box to the user - "You can connect to
>these services - which do you want to connect to?" (where the "services"
>are the groups of servers that have offered addresses; I tried to write
>the dialog in English rather than talking about "offered addresses").

That would be friendly for sure.

>What is the mapping from a server prefix to some human-readable name?

Open discussion.  I don't know off the top of my head.

thanks
/jim


From dhcp-v6@bucknell.edu  Wed Dec  1 23:44:11 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id XAA02410
	for <dhc-archive@odin.ietf.org>; Wed, 1 Dec 1999 23:44:11 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id XAA26233;
	Wed, 1 Dec 1999 23:39:55 -0500 (EST)
Date: Wed, 1 Dec 1999 23:39:55 -0500 (EST)
Message-Id: <199912020416.XAA0000025377@quarry.zk3.dec.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Jim Bound <bound@zk3.dec.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.


>>>[...]it may
>>>be that we need to clarify that a client needs to send multiple Solicit
>>>messages, one to each potential DHCP service, to find the highest
>>>preference server from each DHCP service.  
>>
>>The solicit is a multicast sendmsg() for a link (subnet) then the client
>>will hear from multiple servers, which as you pointed out in other mail
>>we need to clarify for the multiple server case.

>OK, I guess we need to agree on an abstraction for a "DHCP service" and define
>a mechanism for identifying services.  Then, we need to fix 5.3 to state
>that a client applies the server selection algorithm to each service and
>can contact one appropriate server from each service.

Yep.  If we do that first the next step I think will be easier too.
Plus if we do this right we should not have to mess with the other
messages they will work idempotently and if they don't we can fix them
but right now I think they do but will check for sure.

So do you have a suggestion on the abstraction?

thanks
/jim


From dhcp-v6@bucknell.edu  Thu Dec  2 05:40:23 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA09149
	for <dhc-archive@odin.ietf.org>; Thu, 2 Dec 1999 05:40:23 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id FAA00658;
	Thu, 2 Dec 1999 05:37:59 -0500 (EST)
Date: Thu, 2 Dec 1999 05:37:59 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.944106870.9708.nordmark@jurassic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Erik Nordmark <Erik.Nordmark@eng.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0


> How about if we say that a router advertisement with the 'M' bit
> set means that all prefixes advertised in the extensions are for
> managed links?

As far as I can tell that would mean a change to RFC 2462 - which is a draft
standard. Do we really want to change that spec and make it go back to
proposed standard?

If you (and the WG) feels there is a strong need for the routers to
be able to tell the hosts which prefixes the host can/should ask the
dhcp server about, I think it would be better to build that as a backwards
compatible extension to the router advertisement.
But before we go down that path I'd like to be convinced that such
a feature is necessary.

> If a router has some managed links and some unmanaged links,
> then it needs to send out two different router advertisements,
> one for the managed links, and one for the unmanaged links.

Sure, routers already send different router advertisements on different
links since they need to contain different prefixes.

> If more ambitious suggestions were in order, I might suggest
> the creation of an 'M' flag from the reserved flags of the Prefix
> Information option.

That would be one possible way to do this in a backwards compatible way.
But as I said, I don't see the need for this.

   Erik



From dhcp-v6@bucknell.edu  Thu Dec  2 05:40:43 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA09327
	for <dhc-archive@odin.ietf.org>; Thu, 2 Dec 1999 05:40:42 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id FAA10345;
	Thu, 2 Dec 1999 05:38:44 -0500 (EST)
Date: Thu, 2 Dec 1999 05:38:44 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.944122748.17190.nordmark@jurassic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Erik Nordmark <Erik.Nordmark@eng.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DISCUSSION: Use of Address Extensions with DHCPv6
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0


> Semantics of messages:
> 
> 1.  Client sends DHCPv6 Request with no address extension. 
> 
> 1a. Server responds with all addresses available to the client 
>     if the sys admin has defined the server can do this.
> 
> 1b. If the server gets a second message from the client with no
>     address extension it resends the client addresses it gave 
>     them previously cumultatively (see below).
> 
> This would be the default behavior for DHCPv6 and how clients request
> addresses.  The reason there is a sys admin decision is that the network
> admin folks may have decided to use stateless addrconf and dhcpv6 is
> only being used to provide non-address parameters to the clients or to
> update DNS, etc.. 

Jim,

These examples are great trying to nail down the details!

One comment. If a adminstrator wants to use stateless on some links and
dhcp on other links the above seems to be error prone.
Both the stateless boxes asking dhcp for non-address information
and the stateful boxes (asking dhcp for addresses) will send
the same request: one without any address extension.
Thus should the admin get things wrong there will be some confusion.
And the fact that the stateless/managed knob is configured in the routers
sending router advertisements and the "which nodes/links should get addresses"
is configured in the dhcp server, means that misconfiguration will happen.

This *if* the WG concludes that they want different ways to ask for addresses
(as out specify with #1, #2, #3) I think it would be useful to have a
specially marked address extension (for the default set) in #1.
That way a stateless client which asks for non-address info will never
be surprised by receiving address extensions from the server.

  Erik




From dhcp-v6@bucknell.edu  Thu Dec  2 05:40:44 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA09341
	for <dhc-archive@odin.ietf.org>; Thu, 2 Dec 1999 05:40:43 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id FAA23059;
	Thu, 2 Dec 1999 05:38:53 -0500 (EST)
Date: Thu, 2 Dec 1999 05:38:53 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.944123880.2214.nordmark@jurassic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Erik Nordmark <Erik.Nordmark@eng.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0


Charlie Perkins wrote:

> A router advertises a prefix, and its willingness to be a default
> router. 

Correct.

> I infer that it wants to be a default router for that
> prefix.

Incorrect. It is a default router for any IPv6 address.

Even though there are about 6000 lines in RFC 2461 you managed to read
something between those lines that has never been the intent. I'm impressed :-)

  Erik


>  I suspect there would be cases where a router would want
> to be a default router for some prefix(es) and not for others.
> This would be especially true for the multihoming case.  In such
> cases, I reckon distinct advertisements would be needed.
> 
> Or maybe I am confused.
> 
> Regards,
> Charlie P.




From dhcp-v6@bucknell.edu  Thu Dec  2 07:45:56 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA13469
	for <dhc-archive@odin.ietf.org>; Thu, 2 Dec 1999 07:45:55 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA16671;
	Thu, 2 Dec 1999 07:43:24 -0500 (EST)
Date: Thu, 2 Dec 1999 07:43:24 -0500 (EST)
Message-Id: <199912021242.HAA06082@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: Summary: DC meeting question #1
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 06:49 PM 11/30/99 -0500, Michael W. Carney wrote:
>The discussion clearly showed that support for releasable resources was 
>something that the WG wanted. That said, what would be the semantics
>that would resolve the issues raised during the discussion of this
>question? It is this point I'm attempting to address.

Mike - Which discussion are you referring to, here?  IMO, we initiated
the discussion of releasable resources on the mailing to explore the
complexity of adding support to DHCPv6 for releasable resources, so we
could make an informed engineering decision about whether or not to
incorporate releasable resources into the spec.  We have had a good
discussion about how releasable resources might work and how they might
be supported in DHCPv6, but I don't think we've had an expression of
consensus about whether we want to go ahead and add releasable resources
to the spec.

- Ralph




From dhcp-v6@bucknell.edu  Thu Dec  2 20:15:00 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA29783
	for <dhc-archive@odin.ietf.org>; Thu, 2 Dec 1999 20:14:59 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA22191;
	Thu, 2 Dec 1999 20:07:14 -0500 (EST)
Date: Thu, 2 Dec 1999 20:07:14 -0500 (EST)
Message-Id: <3847152D.ADDDAC3C@east.sun.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Michael W. Carney" <michael.carney@east.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: Summary: DC meeting question #1
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.7 [en] (X11; I; Linux 2.2.5-22 i686)
Content-Transfer-Encoding: 7bit

Ralph Droms wrote:
> 
> At 06:49 PM 11/30/99 -0500, Michael W. Carney wrote:
> >The discussion clearly showed that support for releasable resources was
> >something that the WG wanted. That said, what would be the semantics
> >that would resolve the issues raised during the discussion of this
> >question? It is this point I'm attempting to address.
> 
> Mike - Which discussion are you referring to, here?  IMO, we initiated
> the discussion of releasable resources on the mailing to explore the
> complexity of adding support to DHCPv6 for releasable resources, so we
> could make an informed engineering decision about whether or not to
> incorporate releasable resources into the spec.  We have had a good
> discussion about how releasable resources might work and how they might
> be supported in DHCPv6, but I don't think we've had an expression of
> consensus about whether we want to go ahead and add releasable resources
> to the spec.

I fear I may have jumped ahead alittle. You are right. I believe are
close to having the semantics completely defined such that general
support for releasable resources is possible.

Those semantics are:

Either one of the following:

a) Extensions all have a "generic" header portion of which there are
bits which are present in all extensions. One such bit would indicate
that the resource carried in the extension is releasable, and that there
are two fields in the extension (a'la lifetimes in the IP address
extension) which represent the "preferred" and "absolute" lifetimes
represented in seconds relative to the transaction. (now).

b) A range of extensions is carved out and used for releasable
extensions. The fact that the extension is releasable is a
characteristic of the extension's code value.

Client implementations (scheme a or b)
 
	- Must request releasable resources - via extension request extension
in solicit.

	- Will track owner of such resources and SHOULD release them to the
allocating server, and MAY release them to another server if the
allocating server is unreachable.

	- A client will attempt to update the lifetimes on the releasable
resource if necessary. However, if the lifetimes expire, the resource
must be considered no longer usable. There's an interesting point -
since the DHCP client won't deal with most releasable resources, it will
not have control over whether or not a resource continues to be used
when the lifetime expires, whereas it certainly has control over IP
addresses acquired through DHCP.

	- A releasable resources can be reconfigured (again, the client
probably won't have the necessary control of the resource in order to
prevent its use, with the current exception of IP addresses).

Server implementations (scheme a or b)

	- Must have a generic mechanism for dealing with releasable resources.

	- A server will only advertise releasable resources if the client
requests it to do so in the Solicit.

	- A server controls a client's access to releasable resources based
upon the lifetimes associated with these resources. 


So I think this is mostly it in summary. The potential inability of the
client to enforce access to expired releasable resources is an issue.

If folks believe I've covered the required semantics / caveats of
general releasable resources, the question is: Do we try to specify
general releasable resources now, or should we defer till later if we
find them useful?



> 
> - Ralph

-- 

Mike Carney


From dhcp-v6@bucknell.edu  Thu Dec  2 20:15:02 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA29778
	for <dhc-archive@odin.ietf.org>; Thu, 2 Dec 1999 20:14:58 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA14863;
	Thu, 2 Dec 1999 20:05:41 -0500 (EST)
Date: Thu, 2 Dec 1999 20:05:41 -0500 (EST)
Message-Id: <38470F67.1DA00A87@east.sun.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Michael W. Carney" <michael.carney@east.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: Summary: DC meeting question #1
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.7 [en] (X11; I; Linux 2.2.5-22 i686)
Content-Transfer-Encoding: 7bit

"Charles E. Perkins" wrote:
> 
> Hello Mike,
> 
> I hope you don't mind if I put spaces in different places in your
> reply.  It helped me to read your note better, and to read my responses
> to your note better too.

Great, thanks for your comments.

> 
> > 1) General Form
> >
> >     *   Releasable resource extensions need to have a general form so that
> >         clients/servers can manage them in a generic fashion.
> 
> Sounds good.
> 
> >     *   There is a generic bit in the extensions header which is set if the
> >         extension is carrying a releasable resource. When this bit is set,
> >         there are two 32 bit fields present in the extension which represent
> >         the relative times in seconds when the client should start attempting
> >         to extend the lease on the resource, and when the client must stop
> >         using the resource if it is unable to extend it.
> 
> In Minneapolis, the suggestion was made that one range of extension
> types was releasable, and that the rest of the extension types were
> nonreleasable.  There seemed to be some general approval for this idea;
> I don't remember who proposed it.
> 
> If the extensions from 32767-65535 were reserved for releasable
> resources, that would work.  There could be a section in the
> document prescribing a format for these extensions, but in general
> I think that clients that ask for releasable resources can be
> expected to know a lot about those resources.  Thus, working out
> a general format for these future extensions may not be overwhelmingly
> necessary right now.

Yes, I remember this suggestion. I'm fine with reserving some range of
extension codes for releasable resources.

> 
> >     *   Since extensions carrying releasable resources have a general
> >         form/semantics, clients/servers do not need to change when new
> >         extensions are defined which are releasable.
> 
> I don't totally disagree with this statement, but as I just
> described it seems likely that consuming a releasable resource
> implies some other systemic changes in the client that are more
> fundamental than any standardized message format.

Not having to change a client when a new extension is drafted was
something that was specifically requested in Minneapolis, specifically
the "releasability" of a resource.

> 
>  are they allocated/requested, and how and to whom are they
> > released?
> >
> >     *   Servers advertise releasable resources. We (WG) need to consider
> >         how releasable resources are advertised.
> 
> If a client really requires a particular releasable resource, we
> should allow it to append an extension to its solicitation, with
> the meaning that it wants an advertisement from a server that can
> offer the resource.

Or use the "extension request extension" (as you pointed out, we need a
better name).


> 
> > 3) Multiple Server Environments
> >
> >     *   Servers advertise available IP address(es) and other releasable
> >         resources, thus clients know what resources are available from which
> >         server.
> 
> Agreed, but in response to specific client requests only.  Otherwise,
> the advertisements could get pretty long.  Case in point, what if a
> server had to advertise every prefix from which it could allocate
> addresses?  Or every disk farm from which it could allocate scratch
> space?  Or ...?

Or request that such resources be advertised via the Solicit.


> 
> > *       Erik raised an interesting point about application layer extensions on
> >         machines multiple interfaces: If you have multiple NTP servers
> >         available from DHCP information from different interfaces, which one
> >         do you select? In our (Sun's) DHCPv4 client implementation, we have
> >         a notion of a "primary" interface, from which application
> >         layer parameters take precedence over those set in other interfaces.
> >         It seems to me that a similar DHCPv6 implementation strategy could
> >         resolve this issue. Charlie suggested that the server preference
> >         value be used to determine preference. His suggestion has the
> >         advantage that preference is set by the administrator. I prefer
> >         the former solution ("primary" interface).
> 
> Perhaps for extensions where it actually matters, the preference could
> be a field of the extension itself.  My original suggestion was just
> on the spur of the moment, and not made with very much real study.
> Letting the client decide is probably O.K. in most instances, I reckon,
> so if it always decides in favor of a "primary interface", no harm is
> likely to be done.

I agree. Erik?

> 
> Regards,
> Charlie P.

-- 

Mike Carney


From dhcp-v6@bucknell.edu  Fri Dec  3 12:00:26 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA06010
	for <dhc-archive@odin.ietf.org>; Fri, 3 Dec 1999 12:00:26 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA13971;
	Fri, 3 Dec 1999 11:51:06 -0500 (EST)
Date: Fri, 3 Dec 1999 11:51:06 -0500 (EST)
Message-Id: <3847ED7C.6669F80B@iprg.nokia.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Charles E. Perkins" <charliep@iprg.nokia.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: SState machines
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
Content-Transfer-Encoding: 7bit

Ted Lemon wrote:

> Transactions are ephemeral, so I don't see how you can associate state
> machines with them.   Am I misunderstanding what you are saying here?

No, you hit the nail right on the head.  That's exactly why the state
machine should be associated with the transaction, and that is what
was the problem when I had trouble with state machines.  It was that
they were too permanent, having too many global effects.

I like it that when a transaction is done, the state machine managing
the control messages associated with that transaction would disappear.

>                                     I suppose that given that in
> DHCPv6 the server can initiate a transaction, the transaction ID by
> itself couldn't be used to associate a given transaction to a given
> state machine.

But if the state machine is associated with the transaction ID, and
the transaction is completed, the state machine should be completed
(or else be considered as resting in the final, terminal state).

Regards,
Charlie P.


From dhcp-v4@bucknell.edu  Fri Dec  3 13:32:17 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA24187
	for <dhc-archive@odin.ietf.org>; Fri, 3 Dec 1999 13:32:16 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA22711;
	Fri, 3 Dec 1999 13:30:19 -0500 (EST)
Date: Fri, 3 Dec 1999 13:30:19 -0500 (EST)
Message-Id: <4.2.0.58.19991203120438.022b3580@funnel.cisco.com>
Errors-To: droms@bucknell.edu
Reply-To: kkinnear@cisco.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Kim Kinnear <kkinnear@cisco.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Failover conference calls in December
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 


Just note on the failover conference call.

I have received enough people interested in having a failover
conference call to discuss the failover draft, and so I have
scheduled a call for Tuesday, December 7, 1999, from 12-3pm EST.

If you responded to me with interest, you should have already
received email with the conference call connections details.

I sent the original announcement out on November 29, this last
Monday.  If you wish to attend this call and have not already
responded to me, please let me know.

Cheers -- Kim



From dhcp-v4@bucknell.edu  Tue Dec  7 10:38:07 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA14723
	for <dhc-archive@odin.ietf.org>; Tue, 7 Dec 1999 10:38:06 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA13175;
	Tue, 7 Dec 1999 10:36:57 -0500 (EST)
Date: Tue, 7 Dec 1999 10:36:57 -0500 (EST)
Message-Id: <199912071528.JAA00856@symphony.cig.mot.com>
Errors-To: droms@bucknell.edu
Reply-To: nakfour@cig.mot.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Juana E Nakfour <nakfour@cig.mot.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: DHCP over IP/ATM
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: TEXT/PLAIN; charset=US-ASCII
Mime-Version: 1.0

Hello everyone,

 currently we are researching the possiblity of running DHCP over an
IP/ATM network (1577). My question is does DHCP work in such a
network? (DHCP assumes an underlying broadcasting network plus uses
the hardware address as a unique identifier for the client) 
Any help on this subject matter will be greatly appreciated it. Plus
if someone could refer me to any documents covering this topic. Thanks
a lot !!

Cheers !!

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Juana Nakfour                         
Software Eng.                         
BedRock Grp.                          
Motorola                              
(847) 435 8925

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~



From dhcp-v4@bucknell.edu  Tue Dec  7 17:39:59 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA09254
	for <dhc-archive@odin.ietf.org>; Tue, 7 Dec 1999 17:39:59 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id RAA14663;
	Tue, 7 Dec 1999 17:39:01 -0500 (EST)
Date: Tue, 7 Dec 1999 17:39:01 -0500 (EST)
Message-Id: <4.2.0.58.19991207155123.017e7d80@funnel.cisco.com>
Errors-To: droms@bucknell.edu
Reply-To: kkinnear@cisco.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Kim Kinnear <kkinnear@cisco.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Minutes from Dec 7 Failover conference call
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 


Minutes from Failover conference call, held
December 7, 1999 12-2pm EST.

Kim Kinnear, kkinnear@cisco.com

----------------------------------------

NOTE:  There will NOT be a conference
call this Thursday, December 9!

-----------------------------------------

The agenda turned out to be:

1. Security and export controls
2. SASL investigation 
3. Timestamp format - should it be NTP
4. What changed for the -05 draft
5. Document organization
6. POOLREQ -- is this a good technique
7. PARTNER-DOWN and MCLT wait
8. Where do we go from here, short and long term
9. Bulking and group acking

Thanks to everyone who attended:

David Erdmann     Nortel Networks     EST   derdmann@nortelnetworks.com
Steve Gonczi      Process Software    EST   gonczi@process.com
Richard Jones     Checkpoint Software PST   richardj@metaip.checkpoint.com
Kim Kinnear       Cisco Systems       EST   kkinnear@cisco.com
Ted Lemon         Internet Engines    EST   mellon@isc.org
Chris Petrick     Nortel Networks     CST   petrick@nortelnetworks.com
Greg Rabil        Lucent Technologies EST   agrabil@quadritek.com
Bernie Volz       Process Software    EST   volz@process.com


Notes on each topic are below:


1. Security and export controls
------------------------------- 

David Erdmann brought up the concern that with TLS we might get
into export control problems.  Bernie Volz said that we'd picked that
particular cipher because SMTP and IMAP were using it, so
presumably it was "ok", but that he'd certainly check on it.

2. SASL investigation 
--------------------- 

David also wondered where the SASL investigation was, and Ted
Lemon said that he hadn't had time to look into it yet, but he
was going to in a week or so.  Richard Johnson said that he was
interested in this as well.  

3. Timestamp format - should it be NTP
-------------------------------------- 

Kim Kinnear reported that someone had suggested using the NTP
64-bit time format for times for failover.  He did a bit of
research, and discovered that the NTP 64-bit time is 64-bits of
fixed point number.  It is 32 bits to the left of the fixed
point, and 32 bits to the right.  It is based on seconds since
1900 (not 1970), and even though it is unsigned, it expires in
2036.

Kim proposed *not* using the NTP time format, but rather using a
32 bit *unsigned* time format, similar to the ANSI time_t format
for the low order 31 bits, but unsigned.  This will not expire
until 2106 or so.  The rest of the folks agreed.

Richard had pointed out in his review that network byte order
wasn't specified for the time options, and Kim said that in the
next revision he would have a section on time, and reference all
times to that section to explain their format.

4. What changed for the -05 draft
---------------------------------

Kim gave a quick list of what changed in the -05 draft
from the -04 draft.  This list includes:

o new DDNS text from Mark Stapp
o removed load balancing per contributions by Steve Gonczi
o updated security section with new text from Bernie Volz
o added new DHCP client "option buckets" 6.2.27-28
o added new RESOLUTION-INTERRUPTED state
o new section on binding-status state transitions 5.10
o new section on conflict resolution 5.10.1
o new section on reservations 5.12
o new section on dynamic BOOTP 5.13
o put the time in every packet, not in option
o more info on DHCP protocol requirements
o clarified NAK behavior
o clarified how to select MCLT 5.14
o added STARTUP state to section 5, 5.8
o clarified which BNDUPD part of UPDDONE
o changed MAC address to client hardware address
o changed "packet" to "message"
o removed RELEASED-GRACE and EXPIRED-GRACE binding-statuses


5. Document organization
------------------------ 

Richard proposed partitioning the draft into a strict protocol
section and then an explanations and elucidations section.  This
met with some approval, until Kim voiced a concern that yet
another reorganization without tightening up what we've already
got would *look* like progress without perhaps *being* progress.

Kim committed to another edit in an attempt to tighten up what we
already have, with the goal to lean it down and, while increasing
clarity, remove about 20 pages.

We all agreed that we'd look at it after this tightening up, and
if we needed to reorganize it, we'd do it then.

Kim's tasks for the next revision are:

o relate the NAK errors to the rest of the draft
o fix up the times (from #3 above)
o integrate any security changes from Bernie, Ted, Richard
o reorganize conflict resolution into tabular form
o ensure that you can't do both TLS and simple message digest 
o bulking
o add text for PARTNER-DOWN (see #7 below)
o include test in 7.1.1 about mapping both directions
o include Richards comments
o include any other comments he receives
o give entire draft editorial review

Please review the -05 draft and send Kim corrections
and suggestions: kkinnear@cisco.com

6. POOLREQ -- is this a good technique
-------------------------------------- 

Steve Gonczi voiced a concern about the POOLREQ message, in light
of the new loadbalancing approach first adopted in the -04 draft.
Having the primary server send addresses to the secondary server
is nice if the secondary server isn't doing much, but if it is a
major player with DHCP clients, then sometimes the primary server
might need some of the secondary server's IP addresses.

Bernie indicated that the primary server can always ask for them
back from the secondary server if it needs them using BNDUPD
messages.

We discussed for a while allowing the primary server to send
POOLREQ messages to the secondary server, or dropping the POOLREQ
messages altogether and letting each server attempt to pull its
addresses from its partner.  Kim brought up a red-herring about
differing configurations, which in retrospect isn't really an
issue, since so many configuration issues must be the same that
one more doesn't mean much of anything.

At the end of the discussion we didn't agree to changes, but I
believe that Steve said that he might bring it up on this list
for further discussion.

We did agree to add words to suggest that when an address
goes FREE from a secondary server, that the primary server
should probably send it right back to the secondary server.
Of course, not every server will track this because once
an IP address is leased, there is no reason to remember which
server once had it available.

7. PARTNER-DOWN and MCLT wait
-----------------------------

Bernie brought up the issue that, when entering PARTNER-DOWN
state via command, you need to wait the MCLT from when the
partner server went down before using that server's free IP
addresses (and in the absence of any better information, that
time is probably when you got the command).  However, if you go
into PARTNER-DOWN because of a timer that went off in
COMMUNICATIONS-INTERRUPTED state, then you might presume that the
MCLT has already expired for those free addresses.

Kim pointed out that there was another MCLT wait in PARTNER-DOWN,
that from when a lease expires.  That MCLT wait *must* exist in
any case.

Bernie proposed a new state, but we ultimately agreed that we'd
just document that one could assume the MCLT had run out in the
automatic transition to PARTNER-DOWN.

8. Where do we go from here, short and long term
------------------------------------------------

The current plan is:

o produce -06 draft during January
o another conference call late Jan/early Feb
o produce -07 draft prior to Adelaide
o produce -08 draft post Adelaide, last call candidate
o produce -09 draft with changes required for last call

Obviously the last two items are somewhat speculative,
but they are a goal.

9. Bulking and group acking
--------------------------- 

Steve also brought up the issue of bulking, and the way in which
we might ACK.  Kim mentioned that we need to get bulking in the
next revision if we are going to do it at all.  Steve said since
we are using TCP, what about acking only the "latest" BNDUPD, and
let that be an ACK for all previous BNDUPD's.

Kim said that the big performance win with bulking was because
the database flush could be held until the end of all of the
database updates in a bulked packet.  He suggested that the same
could be done with ACK's, but it is harder.

Ted pointed out that to do the "ACK only the last" approach
requires an ordered list of ACK's be kept, and Kim said that some
implementations might not want to keep such a list.  Then Kim
pointed out that a new implementation might not even keep
outstanding BNDUPD messages around at all, but might well just
rely on the partner to send back the BNDACK messages, and update
the lease database based solely on the BNDACK messages.  This is
possible since retransmission is not required.

We ended up leaving it the way it was, with Steve probably going
to bring it up on the list for further discussion.

----------------- end of minutes ---------------------




From dhcp-v4@bucknell.edu  Tue Dec  7 18:29:59 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA02757
	for <dhc-archive@odin.ietf.org>; Tue, 7 Dec 1999 18:29:56 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id SAA26222;
	Tue, 7 Dec 1999 18:28:50 -0500 (EST)
Date: Tue, 7 Dec 1999 18:28:50 -0500 (EST)
Message-Id: <v04220802b4734196e9b1@[139.132.83.10]>
Errors-To: droms@bucknell.edu
Reply-To: pmlegge@deakin.edu.au
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Philip Legge <pmlegge@deakin.edu.au>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: DHCP over IP/ATM
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii" ; format="flowed"
Mime-Version: 1.0

>Date: Tue, 7 Dec 1999 09:28:01 -0600 (CST)
>From: Juana E Nakfour <nakfour@cig.mot.com>
>Subject: DHCP over IP/ATM
>
>  currently we are researching the possiblity of running DHCP over an
>IP/ATM network (1577). My question is does DHCP work in such a
>network? (DHCP assumes an underlying broadcasting network plus uses
>the hardware address as a unique identifier for the client)
>Any help on this subject matter will be greatly appreciated it. Plus
>if someone could refer me to any documents covering this topic. Thanks
>a lot !!

Hi Juana,

DHCP does work well over ATM but there seem to be interesting 
anomalies, so it would be helpful if any documents regarding this are 
brought to the attention of the list.

At Deakin we are using switched ethernet with ATM implemented between 
switches and router, which works fine except when a client (usually a 
laptop) changes subnet and is booted. When the laptop is booted and 
issues a DHCPREQUEST, ATM is able to deliver the message to the local 
DHCP server; the DHCP server does issue DHCPNAK as it recognises the 
client has changed subnet, however ATM is unable to deliver this 
message back to the client. Without DHCPNAK the client continues to 
use it's previous lease until expiry (and thus can't use the network 
with TCP/IP misconfigured).

Since the problem doesn't occur if the previous lease address expires 
or is released, we have a variety of methods to overcome this - use 
of scripts to force DHCPRELEASE on shutdown or startup, or the 
graphical utilities for Microsoft DHCP clients (winipcfg and 
wntipcfg, the latter is found on the NT4Wks Resource Kit CD).

BTW, DHCP doesn't assume the hardware address to always be a unique 
indentifier, however the various Microsoft clients usually do!

Cheers, Philip

--
Philip Legge <pmlegge@deakin.edu.au> Deakin Uni - Business & Law- IT Support


From dhcp-v4@bucknell.edu  Wed Dec  8 16:43:46 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA05126
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 16:43:44 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA28621;
	Wed, 8 Dec 1999 16:41:46 -0500 (EST)
Date: Wed, 8 Dec 1999 16:41:46 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE7003@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0

Greetings,

I would like to point out a problem and propose a possible solution to the
primary-backup address management scheme in the latest version of the DHC
Failover protocol, which is IMHO slightly broken.
Comments, questions are most welcome!

How it used to work: 

a) When the primary and backup servers are done synchronizing their
databases,
the backup server sends a POOLREQUEST, asking for addresses to replenish its
pool. The primary typically assigns some and sends them over as binding
updates. 

b) When operating in normal mode, the backup server DOES NOT use up any of
these addresses, as it does not server new clients.

c) In communications interrupted mode, the backup server starts using up its
address pool. 

d) Any addresses it has assigned, and freed ( expired or released) revert
back to the ownership of the primary. This happens because the lease
ownership is expressed merely by setting the lease status to "BACKUP". If it
is anything but "BACKUP" the backup server no longer owns it.

e) When the servers re-connect, a) happens again, so life is good.

What is different in the latest draft:

1) Because of load balancing, the backup server now serves new clients from
its pool during normal operation, hence it continually depletes its address
pool. It is virtually guaranteed to run out of addresses during normal
operation,
unless we take steps to prevent this... 
Why? See paragraph d)..

2) The POOLREQUEST /POOLREPLY exchange is relatively expensive on the
primary server side, because the primary
server has to scan its entire lease database to figure out which addresses
(if any) to transfer. 

3) The POOLREQUEST/POOLREPLY exchange should only be used immediately after
the primary has synchronized its database with the backup, so it has a
complete, up-to-date view of the lease database.

Possible solution: 

I) Both servers should flag and keep track of addresses owned by the backup
(in addition to and beyond the lease being in the "BACKUP" state. If a lease
is assigned, i.e. it is "ACTIVE" it still should retain its flag, marking it
as a "BACKUP" lease in BOTH servers.

II) Whenever the backup server detects an expiration, or release of a lease
it owns, it should set the status back to "BACKUP" not "FREE" and should
send an update to the primary with the lease status set to "BACKUP". 

!!!NOTE!!!
The above has nothing to do with my earlier (foolish) proposal of not
allowing a server renew a lease it did not own, if it was marked
free/backup. 


Another issue, loosely related to the previous one:

A) The POOLREQUEST/POOLREPLY exchange should work as an initial pool
arbitration process (pretty much as it works today). The pool percentage is
either a configuration value or possibly a self-tuning parameter (?)

B) To allow the administrator to tune the backup/primary address ratio (say,
based on server load analysis)
we should allow the primary to TAKE AWAY, as well as hand over addresses. 

C) Hence the number of addresses allocated which we report in the POOLREPLY
should either allowed to be
negative as well as positive, or it should be changed to a mere status value
( 1==success, 0==failure)

D) For B) to work, we must make sure the backup server is not serving
clients until the primary has finished transferring the addresses. This is
because it may decide to take away an address, that the backup  could 
have leased in the meantime. 



From dhcp-v6@bucknell.edu  Wed Dec  8 16:49:20 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA08062
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 16:49:18 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA23363;
	Wed, 8 Dec 1999 16:37:11 -0500 (EST)
Date: Wed, 8 Dec 1999 16:37:11 -0500 (EST)
Message-Id: <384EE204.82179E6E@east.sun.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Michael W. Carney" <michael.carney@east.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: WG: general releasable resources?
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.7 [en] (X11; I; Linux 2.2.5-22 i686)
Content-Transfer-Encoding: 7bit


Well, we've talked alot about general releasable resources and what
their semantics might be (summary below).

The question I have for the WG is:

Do we attempt to specify general releasable resources now, or should we
defer their specification till later if we find them useful?

Please respond with your preference.

Thanks!


"Michael W. Carney" wrote:
> 
> Ralph Droms wrote:
> >
> > At 06:49 PM 11/30/99 -0500, Michael W. Carney wrote:
> > >The discussion clearly showed that support for releasable resources was
> > >something that the WG wanted. That said, what would be the semantics
> > >that would resolve the issues raised during the discussion of this
> > >question? It is this point I'm attempting to address.
> >
> > Mike - Which discussion are you referring to, here?  IMO, we initiated
> > the discussion of releasable resources on the mailing to explore the
> > complexity of adding support to DHCPv6 for releasable resources, so we
> > could make an informed engineering decision about whether or not to
> > incorporate releasable resources into the spec.  We have had a good
> > discussion about how releasable resources might work and how they might
> > be supported in DHCPv6, but I don't think we've had an expression of
> > consensus about whether we want to go ahead and add releasable resources
> > to the spec.
> 
> I fear I may have jumped ahead alittle. You are right. I believe are
> close to having the semantics completely defined such that general
> support for releasable resources is possible.
> 
> Those semantics are:
> 
> Either one of the following:
> 
> a) Extensions all have a "generic" header portion of which there are
> bits which are present in all extensions. One such bit would indicate
> that the resource carried in the extension is releasable, and that there
> are two fields in the extension (a'la lifetimes in the IP address
> extension) which represent the "preferred" and "absolute" lifetimes
> represented in seconds relative to the transaction. (now).
> 
> b) A range of extensions is carved out and used for releasable
> extensions. The fact that the extension is releasable is a
> characteristic of the extension's code value.
> 
> Client implementations (scheme a or b)
> 
>         - Must request releasable resources - via extension request extension
> in solicit.
> 
>         - Will track owner of such resources and SHOULD release them to the
> allocating server, and MAY release them to another server if the
> allocating server is unreachable.
> 
>         - A client will attempt to update the lifetimes on the releasable
> resource if necessary. However, if the lifetimes expire, the resource
> must be considered no longer usable. There's an interesting point -
> since the DHCP client won't deal with most releasable resources, it will
> not have control over whether or not a resource continues to be used
> when the lifetime expires, whereas it certainly has control over IP
> addresses acquired through DHCP.
> 
>         - A releasable resources can be reconfigured (again, the client
> probably won't have the necessary control of the resource in order to
> prevent its use, with the current exception of IP addresses).
> 
> Server implementations (scheme a or b)
> 
>         - Must have a generic mechanism for dealing with releasable resources.
> 
>         - A server will only advertise releasable resources if the client
> requests it to do so in the Solicit.
> 
>         - A server controls a client's access to releasable resources based
> upon the lifetimes associated with these resources.
> 
> So I think this is mostly it in summary. The potential inability of the
> client to enforce access to expired releasable resources is an issue.
> 
> If folks believe I've covered the required semantics / caveats of
> general releasable resources, the question is: Do we try to specify
> general releasable resources now, or should we defer till later if we
> find them useful?
> 
> >
> > - Ralph
> 
> --
> 
> Mike Carney

-- 

Mike Carney


From dhcp-v6@bucknell.edu  Wed Dec  8 18:11:23 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA20308
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 18:11:22 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id SAA13994;
	Wed, 8 Dec 1999 18:05:11 -0500 (EST)
Date: Wed, 8 Dec 1999 18:05:11 -0500 (EST)
Message-Id: <4D0A23B3F74DD111ACCD00805F31D8101CA21814@RED-MSG-50>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Richard Draves <richdr@microsoft.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: RE: general releasable resources?
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
X-Mailer: Internet Mail Service (5.5.2650.21)

> Well, we've talked alot about general releasable resources and what
> their semantics might be (summary below).
> 
> The question I have for the WG is:
> 
> Do we attempt to specify general releasable resources now, or 
> should we
> defer their specification till later if we find them useful?
> 
> Please respond with your preference.

Defer their specification until later.

Thanks,
Rich


From dhcp-v6@bucknell.edu  Wed Dec  8 18:11:35 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA20417
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 18:11:35 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id SAA20779;
	Wed, 8 Dec 1999 18:05:41 -0500 (EST)
Date: Wed, 8 Dec 1999 18:05:41 -0500 (EST)
Message-Id: <21839C593E75D211BF7800805FCCD026EAAEAA@msgsrv03.srv.PacBell.COM>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "HIBBS, BARR (PB)" <RBHIBBS@msg.pacbell.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: RE: general releasable resources?
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain
X-Mailer: Internet Mail Service (5.5.2448.0)
Mime-Version: 1.0


Mike, Ralph, et al.,

Inasmuch as we've only determined one releasable resource, the IP address,
to date, I'm in favor of making a first pass at the general solution now:
the existence of a versioning mechanism makes it possible for us to replace
(or augment) such a mechanism in the future if our operating experience
isn't great.

I also favor approach 'A' with an extension header that identifies the
resource as being releasable.

However, I think there is one issue we must resolve first:  why must a
client request a releasable resource, but not be required to explicitly
request non-releasable ones?  If some extensions are intended to always be
supplied by a server, are these specified in one of the drafts?  I actually
favor a scheme where clients explicitly request all extensions that they
desire, with the server supplying any that are required because of [other]
protocol requirements or local site policy.  I just want to see the language
about what is and which ones are automatically provided and which ones must
be requested to be unambiguous in the RFCs.

--Barr





From dhcp-v4@bucknell.edu  Wed Dec  8 19:01:06 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA15560
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 19:01:05 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id SAA00553;
	Wed, 8 Dec 1999 18:59:38 -0500 (EST)
Date: Wed, 8 Dec 1999 18:59:38 -0500 (EST)
Message-Id: <384EE71A.AD69200A@metaip.checkpoint.com>
Errors-To: droms@bucknell.edu
Reply-To: richardj@metaip.checkpoint.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Richard Jones <richardj@metaip.checkpoint.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
Content-Transfer-Encoding: 7bit

My first take on this is that the proposal muddies the relationship between
primary and secondary in the failover protocol, all in the service of load
balancing - which I believe we've pretty much decided is peripheral
to the function of failover.

I'm in favor of maintaining the priority of the primary server. I don't think
POOLREQ functions need to be expensive, if the primary is implemented
to know how many addresses the secondary has and how many it's allowed
(or configured) to have. Maybe that's going to be harder to do than I
think ;-).

I always assumed that POOLREQ/POOLRESP exchanges could happen
any time, and that when they did happen, there would be a period of
extra bind updates and then it would be done for a while. But compared
to the rest of the traffic, it doesn't seem that awful, especially since the
folks who have implemented already seem to feel that wire traffic is
nothing compared to other functions (like stable storage).

If we have the secondary change the state of a lease to BACKUP and
update the primary, what happens in the case of a conflict? Maybe we'd
simply need another reject-reason code, but there may be more to it
than that.

All in all, I don't think the protocol as it stands is broken in this respect.

Regards,
Richard

Steve Gonczi wrote:

> Greetings,
>
> I would like to point out a problem and propose a possible solution to the
> primary-backup address management scheme in the latest version of the DHC
> Failover protocol, which is IMHO slightly broken.
> Comments, questions are most welcome!
>
> Possible solution:
>
> I) Both servers should flag and keep track of addresses owned by the backup
> (in addition to and beyond the lease being in the "BACKUP" state. If a lease
> is assigned, i.e. it is "ACTIVE" it still should retain its flag, marking it
> as a "BACKUP" lease in BOTH servers.
>
> II) Whenever the backup server detects an expiration, or release of a lease
> it owns, it should set the status back to "BACKUP" not "FREE" and should
> send an update to the primary with the lease status set to "BACKUP".
>
> !!!NOTE!!!
> The above has nothing to do with my earlier (foolish) proposal of not
> allowing a server renew a lease it did not own, if it was marked
> free/backup.
>
> Another issue, loosely related to the previous one:
>
> A) The POOLREQUEST/POOLREPLY exchange should work as an initial pool
> arbitration process (pretty much as it works today). The pool percentage is
> either a configuration value or possibly a self-tuning parameter (?)
>
> B) To allow the administrator to tune the backup/primary address ratio (say,
> based on server load analysis)
> we should allow the primary to TAKE AWAY, as well as hand over addresses.
>
> C) Hence the number of addresses allocated which we report in the POOLREPLY
> should either allowed to be
> negative as well as positive, or it should be changed to a mere status value
> ( 1==success, 0==failure)
>
> D) For B) to work, we must make sure the backup server is not serving
> clients until the primary has finished transferring the addresses. This is
> because it may decide to take away an address, that the backup  could
> have leased in the meantime.



From dhcp-v4@bucknell.edu  Wed Dec  8 19:04:42 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA17365
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 19:04:39 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA05744;
	Wed, 8 Dec 1999 19:03:27 -0500 (EST)
Date: Wed, 8 Dec 1999 19:03:27 -0500 (EST)
Message-Id: <199912082322.SAA24180@agen.ufl.edu>
Errors-To: droms@bucknell.edu
Reply-To: wally@water.agen.ufl.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Wally Hunter <wally@water.agen.ufl.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Subscribe
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
X-Mailer: dtmail 1.3.0 @(#)CDE Version 1.3.4 SunOS 5.7 sun4m sparc 
Content-Type: TEXT/plain; charset=us-ascii
Mime-Version: 1.0

Subscribe



From dhcp-v4@bucknell.edu  Wed Dec  8 19:47:57 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA08677
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 19:47:57 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA24803;
	Wed, 8 Dec 1999 19:47:13 -0500 (EST)
Date: Wed, 8 Dec 1999 19:47:13 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE7004@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0

Richard,

Thanks for your quick review and response.

>My first take on this is that the proposal muddies the relationship between
>primary and secondary in the failover protocol, all in the service of load
>balancing - which I believe we've pretty much decided is peripheral
>to the function of failover.

I believe (and the protocol authors will probably agree) that making the
protocol 
more symmetrical is a good thing. As I recall, the idea has been around for
a
while. The first step in this direction was to allow the backup server to
renew.
leases while in normal mode.


>I'm in favor of maintaining the priority of the primary server. 

What advantage do you see in doing so?

>I don't think POOLREQ functions need to be expensive, if the primary is
implemented
>to know how many addresses the secondary has and how many it's allowed
>(or configured) to have. Maybe that's going to be harder to do than I
>think ;-).

I am afraid it is a bit harder. New trends in server operation introduce
complexities like
arbitrary lease pools, (with different settings and privileges), which leads
to the primary 
having to leaf through all of them, to hand over the prescribed percantage
of each.

>I always assumed that POOLREQ/POOLRESP exchanges could happen
>any time, and that when they did happen, there would be a period of
>extra bind updates and then it would be done for a while. But compared
>to the rest of the traffic, it doesn't seem that awful, especially since
the
>folks who have implemented already seem to feel that wire traffic is
>nothing compared to other functions (like stable storage).

It all depends on the kinds of trafic you are handling. If you want carrier
class performance, 
wire traffic does become an issue.

>If we have the secondary change the state of a lease to BACKUP and
>update the primary, what happens in the case of a conflict? Maybe we'd
>simply need another reject-reason code, but there may be more to it
>than that.

I would very much like to avoid cycling the backup server's expiring leases
through 
the primary if it can be avoided, for reasons of efficiency, and keeping the
same 
lease with the same server.

If someone can come up with a good reason why we have to do this, I would
very 
much like to hear it.

sG


From dhcp-v4@bucknell.edu  Wed Dec  8 19:50:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA09993
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 19:50:39 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA25145;
	Wed, 8 Dec 1999 19:50:21 -0500 (EST)
Date: Wed, 8 Dec 1999 19:50:21 -0500 (EST)
Message-Id: <199912090006.QAA01292@grosse.manhattan.fugue.com>
Errors-To: droms@bucknell.edu
Reply-To: mellon@isc.org
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ted Lemon <mellon@isc.org>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4


> My first take on this is that the proposal muddies the relationship between
> primary and secondary in the failover protocol, all in the service of load
> balancing - which I believe we've pretty much decided is peripheral
> to the function of failover.

Wow, I hadn't decided that.   I think that load balancing makes the
failover protocol much more useful.

Other than that, I didn't see anything wrong with what you said, but I
haven't finished implementing yet, so I reserve the right to comment
further... :')

			       _MelloN_


From dhcp-v4@bucknell.edu  Wed Dec  8 19:53:38 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA11293
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 19:53:36 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA04554;
	Wed, 8 Dec 1999 19:53:17 -0500 (EST)
Date: Wed, 8 Dec 1999 19:53:17 -0500 (EST)
Message-Id: <384EF8EE.35BF68A@metaip.checkpoint.com>
Errors-To: droms@bucknell.edu
Reply-To: richardj@metaip.checkpoint.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Richard Jones <richardj@metaip.checkpoint.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
Content-Transfer-Encoding: 7bit

Ted Lemon wrote:

> > My first take on this is that the proposal muddies the relationship between
> > primary and secondary in the failover protocol, all in the service of load
> > balancing - which I believe we've pretty much decided is peripheral
> > to the function of failover.
>
> Wow, I hadn't decided that.   I think that load balancing makes the
> failover protocol much more useful.

I didn't mean to speak out of turn. I guess what I was trying to say
was that since we decided that it was appropriate to put the load
balancing functionality in another draft, it was less of a candidate
for driving  protocol details. Sounds like that would not meet with
general agreement.


>
>
> Other than that, I didn't see anything wrong with what you said, but I
> haven't finished implementing yet, so I reserve the right to comment
> further... :')
>
>                                _MelloN_

By all means! A few weeks from now I'll probably have some
different ideas, too ;-).

Regards,
Richard




From dhcp-v4@bucknell.edu  Wed Dec  8 19:56:27 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA12623
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 19:56:26 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA12845;
	Wed, 8 Dec 1999 19:56:01 -0500 (EST)
Date: Wed, 8 Dec 1999 19:56:01 -0500 (EST)
Message-Id: <199912090034.QAA01655@grosse.manhattan.fugue.com>
Errors-To: droms@bucknell.edu
Reply-To: mellon@isc.org
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ted Lemon <mellon@isc.org>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4


> I guess what I was trying to say was that since we decided that it
> was appropriate to put the load balancing functionality in another
> draft, it was less of a candidate for driving protocol
> details. Sounds like that would not meet with general agreement.

Nope.   We just moved the technical description of load balancing out
of the draft.   It's still required by the draft, unless I am much
mistaken.   So it's as important as it was before the decision to
remove the details from the failover draft was made.

			       _MelloN_


From dhcp-v4@bucknell.edu  Wed Dec  8 20:36:15 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA02542
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 20:36:12 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA11478;
	Wed, 8 Dec 1999 20:35:22 -0500 (EST)
Date: Wed, 8 Dec 1999 20:35:22 -0500 (EST)
Message-Id: <384F01A6.9C3029E0@metaip.checkpoint.com>
Errors-To: droms@bucknell.edu
Reply-To: richardj@metaip.checkpoint.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Richard Jones <richardj@metaip.checkpoint.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
Content-Transfer-Encoding: 7bit



Steve Gonczi wrote:

> Richard,
>
> Thanks for your quick review and response.
>
> I believe (and the protocol authors will probably agree) that making the
> protocol
> more symmetrical is a good thing. As I recall, the idea has been around for
> a
> while. The first step in this direction was to allow the backup server to
> renew.
> leases while in normal mode.

I've heard this in the group's discussions, and in general I'm right with it.
But I  think it's fairly safe to say that this protocol can't be completely
symmetrical, so there may be a point where we have to ask, "How
symmetrical is too much? When does complexity overcome the
advantages of symmetry?" I don't really know how to answer that
at this point, and in fact I'm ready to defer to those folks who have
been working with this for much longer than I. It just seems like a
question worth asking, often ;-).

> >I'm in favor of maintaining the priority of the primary server.
>
> What advantage do you see in doing so?

Perhaps this is just a reflection of gaps in my experience, but it
seems like maintaining the priority of one over the other reduces
the possibility of conflict in general, and makes arbitration simpler.

> >I don't think POOLREQ functions need to be expensive, if the primary is
> implemented
> >to know how many addresses the secondary has and how many it's allowed
> >(or configured) to have. Maybe that's going to be harder to do than I
> >think ;-).
>
> I am afraid it is a bit harder. New trends in server operation introduce
> complexities like
> arbitrary lease pools, (with different settings and privileges), which leads
> to the primary
> having to leaf through all of them, to hand over the prescribed percantage
> of each.

I'm not familiar with arbitrary lease pools. Are they analogous to vendor-
or user-classes? Anyway, I can't defend this point strongly because I
haven't done it yet, but yes, it could (make that *will*) certainly get
complicated.

> >I always assumed that POOLREQ/POOLRESP exchanges could happen
> >any time, and that when they did happen, there would be a period of
> >extra bind updates and then it would be done for a while. But compared
> >to the rest of the traffic, it doesn't seem that awful, especially since
> the
> >folks who have implemented already seem to feel that wire traffic is
> >nothing compared to other functions (like stable storage).
>
> It all depends on the kinds of trafic you are handling. If you want carrier
> class performance,
> wire traffic does become an issue.

Okay.

> >If we have the secondary change the state of a lease to BACKUP and
> >update the primary, what happens in the case of a conflict? Maybe we'd
> >simply need another reject-reason code, but there may be more to it
> >than that.
>
> I would very much like to avoid cycling the backup server's expiring leases
> through
> the primary if it can be avoided, for reasons of efficiency, and keeping the
> same
> lease with the same server.

> If someone can come up with a good reason why we have to do this, I would
> very
> much like to hear it.
>
> sG

It could well be that my reasons above don't stand up to the gains in
efficiency  that Steve's proposal might yield. But I am interested in the question

of peer symmetry versus protocol complexity, so if anyone has something
about that, I'd like to hear it too.

Regards,
Richard




From dhcp-v4@bucknell.edu  Wed Dec  8 20:40:47 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA04861
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 20:40:43 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA18052;
	Wed, 8 Dec 1999 20:40:14 -0500 (EST)
Date: Wed, 8 Dec 1999 20:40:14 -0500 (EST)
Message-Id: <002001bf41e2$4060c000$fe8ac8c6@quadritek.com>
Errors-To: droms@bucknell.edu
Reply-To: agrabil@quadritek.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: agrabil@quadritek.com (Greg Rabil)
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2173.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit

> > I guess what I was trying to say was that since we decided that it
> > was appropriate to put the load balancing functionality in another
> > draft, it was less of a candidate for driving protocol
> > details. Sounds like that would not meet with general agreement.
>
> Nope.   We just moved the technical description of load balancing out
> of the draft.   It's still required by the draft, unless I am much
> mistaken.   So it's as important as it was before the decision to
> remove the details from the failover draft was made.
>
> 			       _MelloN_
>

OK, now I'm confused.  I thought (like Richard) that we'd decided to move
the load-balancing into a separate draft because is was _not_ required for
failover.  Yes, it is nice to have, but I did not realize that we were
requiring it.  If we are, then why publish a separate draft?  Is the
failover draft now dependent upon the load-balancing draft for standard?

Greg



From dhcp-v4@bucknell.edu  Wed Dec  8 21:17:45 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA23089
	for <dhc-archive@odin.ietf.org>; Wed, 8 Dec 1999 21:17:33 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id VAA01281;
	Wed, 8 Dec 1999 21:16:52 -0500 (EST)
Date: Wed, 8 Dec 1999 21:16:52 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE7005@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0

>I'm not familiar with arbitrary lease pools. Are they analogous to vendor-
>or user-classes? Anyway, I can't defend this point strongly because I
>haven't done it yet, but yes, it could (make that *will*) certainly get
>complicated.

They are useful/needed in supporting user/vendor classing. Just a 
way to group arbitrary IP addresses together for whatever purpose.
E.g.: a group of addresses are set up to be allowed through a firewall
and you give a lease from THAT pool if the client matches 
some criteria. The above mentioned addresses form a pool.
Ted has a very nice implementation in his 3.x servers. 
You may want to take a look.

>But I am interested in the question of peer symmetry versus protocol
complexity, so if anyone has something
>about that, I'd like to hear it too.

We need to get the amount of symmetry just right ;-)

sG


From dhcp-v6@bucknell.edu  Thu Dec  9 05:35:31 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA11435
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 05:35:31 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id FAA14538;
	Thu, 9 Dec 1999 05:29:37 -0500 (EST)
Date: Thu, 9 Dec 1999 05:29:37 -0500 (EST)
Message-Id: <199912091022.LAA16625@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 Input from Francis Dupont 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   
   >  - I prefer to have a 32 bit interface index (relative to the agent)
   >    in place of the agent address + prefix-size. I am not convinced
   >    address + prefix-size is enough for all topologies, an interface index
   >    is clearer and easier to use (as a scope ID). This solves the alignement
   >    problem to if link-local addresses are replaced by interface index +
   >    link-local address.
   
   If we go the above path this implies that a client will be identified
   by the ifindex on the relay plus the client's link-local address.

=> there are two different problems:
 - how the server identifies the client (and should the server recognize
   a client using a different relay?)
 - how the relay gets the link to use in a packet from a server.
My concern is for the second problem the only good solution is to have
the interface index because a naked address is not enough.

   >  - the real source of the problem is in RFC 2462 5.5.3 "Router
   >    Advertisement Processing" page 17 where you can find after two paragraphs
   >    about M (managed) and O (other) bits this statement:
   >    "For each Prefix-Information option in the Router Advertisement: a-e)"
   >    obviously the text after this statement and the 5.6 "Configuration
   >    Consistency" section next page (with "hosts accept the union of all
   >    information ..." are misleading in the stateful autoconf context.
   
   What is misleading?

=> the case where the M bit and A flags are set was not clear for me
then I propose to clarify this with a statement explaining that
when the M bit is set then A flags should not be set.
(of course this statement can go into a DHCPv6 document)

   But isn't "stateless mode is available" the same as "one or more prefixes
   have the A bit set"?

=> yes

   Thus I think we already have the necessary knobs.
   
=> I agree, we should only clarify this point in a DHCPv6 document.

Thanks

Francis.Dupont@inria.fr


From dhcp-v6@bucknell.edu  Thu Dec  9 06:05:50 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA26514
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 06:05:50 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id GAA21437;
	Thu, 9 Dec 1999 06:00:15 -0500 (EST)
Date: Thu, 9 Dec 1999 06:00:15 -0500 (EST)
Message-Id: <199912091037.LAA16857@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 Input from Francis Dupont 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   >*If* we agree that a client should be identified as the same client
   >by a server independent of the relay agent it is using on the link
   >then I think the spec needs a slight tweak.
   
   I think we all agreed long ago that a client needs to be identified as
   you very nicely stated, if anyone does not agree I would be surprised.
   
=> then the answer is yes and we should check if the current proposal
covers all the cases...

   >It currently says that the client is identified by <link-local address,
   >agent-address> and I think this should be <link-local address, agent-prefix>.
   >(where agent-prefix could also be called link-prefix since it is the prefix of
   >the client's link that matters).
   
   It should say agent-prefix that was a typo in draft-14 for some reason,
   the authors agree with you, and there are other places where we speak of
   prefix for sure, but this is an edit error in the spec presently.
   
=> the use of the prefix-size was a recurrent weakness of draft-XX.
I hope this discussion will give a better I-D... (:-)

   >In addition to clarifying the scope (site-local vs. global) for
   >the agent-prefix as you point out we also need to make sure that
   >two agents on a link pick the same prefix as agent-prefix,
   >since there can be multiple global prefixes assigned to the link.
   >I think some simple rules (pick the numerically largest prefix)
   >would suffice.
   
   I agree.  We need to add that and the scope issue.
   
=> I believe we have an issue for "off-link" prefixes too because
they are by definition shared between links ?!

Regards

Francis.Dupont@inria.fr


From dhcp-v6@bucknell.edu  Thu Dec  9 06:39:32 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA13745
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 06:39:32 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id GAA08194;
	Thu, 9 Dec 1999 06:33:12 -0500 (EST)
Date: Thu, 9 Dec 1999 06:33:12 -0500 (EST)
Message-Id: <199912091109.MAA27134@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DISCUSSION: Use of Address Extensions with DHCPv6 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   Semantics of messages:
   
   1.  Client sends DHCPv6 Request with no address extension. 
   
=> I have a real concern with this because a client should be
allowed to use DHCPv6 for other things than statefull autoconf,
ie. the semantics of no address extension in a request should be
"I don't want an address".
Another point: #1 is no address extension or no extension at all
(ie. there is a conflict with #3)

   1a. Server responds with all addresses available to the client 
       if the sys admin has defined the server can do this.
   
=> this breaks the one for one releasable resource rule.
Is it a real issue?

   1b. If the server gets a second message from the client with no
       address extension it resends the client addresses it gave 
       them previously cumultatively (see below).
   
=> this supposes the client can be identified in all cases
(this is the only issue I can see).

   This would be the default behavior for DHCPv6 and how clients request
   addresses.  The reason there is a sys admin decision is that the network
   admin folks may have decided to use stateless addrconf and dhcpv6 is
   only being used to provide non-address parameters to the clients or to
   update DNS, etc.. 
   
=> Argh! In fact I believe you'd like to have this #1 behavior ONLY
when the server knows the client is on a "managed" link. In this
context then I'd like to change my concern against the whole #1
to a request for a clarification!

   2b. The client MUST NOT send this through a RELAY-AGENT and the 
       client is assumed to have an address with enough scope to 
       reach the server directly.

=> a stupid (?) question: how you can identify the client in this case?

   3.  Client sends Request with address extension or extensions
       which are not empty.
   
=> I don't really understand the "extensions which are not empty"...
I believe we should have a way in this context to get an extra address
in a prefix, I believe your answer is "form an address in this prefix
and check with the server if this address can be used", this is good
for me.
   
Thanks

Francis.Dupont@inria.fr


From dhcp-v6@bucknell.edu  Thu Dec  9 07:11:22 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA29471
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 07:11:22 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA04710;
	Thu, 9 Dec 1999 07:06:39 -0500 (EST)
Date: Thu, 9 Dec 1999 07:06:39 -0500 (EST)
Message-Id: <199912091146.MAA01388@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DISCUSSION: Use of Address Extensions with DHCPv6 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   Hi Mike,
   
   >Seems to me (after no small amount of consideration) that a server needs to
   >provide more information in its advertisements then we have been
   >proposing (note that providing this information doesn't change the
   >protocol as defined at all). Consider my summary of
   >question #1 - I've used it in responding to this message.
   
=> I'd like to get a clarification about what kind of infos may/should
be in advertisements too.
   
   >#1 goes against the constraint I listed in the Summary: DC meeting
   >question #1 message: Clients must explicitly request releaseable
   >resources.

=> I had the same reaction but now I believe Jim wanted to put #1 in
a special context where the server knows DHCPv6 is used for address
management then #1 simplifies the whole thing.

   >If the server had explicit (through policy) or implicit
   >(dynamic addresses) to offer to the client, it would do so by adding
   >IP address extensions carrying these addresses to the
   >advertisement. The client would request one or all of these IP
   >addreses in its request. If the site is using stateless autoconf, then
   >the server would not have any IP address extensions in its
   >advertisement, and thus the client could not request an IP address.
   
   that would work yes.  same affect.  I can live with this if others can?
   
=> we need to have an exchange with extensions in order to negociate
DNS stuff for instance. If they are in advertisements this will work
(ie. I can live with this, this is in fact one of the solutions for
the multiple address management issue).

   >I'm assuming that this request is in addition to the IP
   >address(es) offered in the advertisement. The client will have
   >knowledge of what networks are available from the advertisement, and
   >thus would need to tell the server which network it would like an IP
   >address on. Thus, an empty IP address extension isn't enough - the
   >client would need to communicate which network.

=> I believe you should do it with the #3, the only difference is
with #3 the client has to build an address in the prefix.

   >At DC, I proposed a change to the IP address extension which would
   >permit the client to identify which network it wants an additional IP
   >address on. Whether or not it would receive an address is up to
   >administrator policy.

=> perhaps this introduces an unneeded complexity? I can live with
the rule that servers give primary addresses (ie. one address in each
prefix for each client on a "managed" link) and clients build and check
(with servers) secondary/extra addresses (ie. an address not in the previous
case).

   >Note that a server could use this modification for advertising
   >network prefixes, without advertising a specific IP address on that
   >network.
   
=> I think new prefixes should be dealt with reconfigure then I can't
see where this change is really needed.

   Again this is an optimization I can live with.
   
=> Jim, I am afraid it is an extra complexity and not an optimization.

   But for this case a client would have to provide an extension for each
   address it wanted.
  
=> this is a good thing because the client wanted extra things then
it needed fine control: my idea is we have to split the address management
is two parts, one for primary addresses and one for secondary addresses.

   Do you agree with this semantic for the protocol?
   
=> we need to write it down again before (and to get more opinions)

   And with the behavior?  This is one of Thomas's issues as a note and I
   would like to resolve that meta issue very clearly this time?
   
=> this issue is important because without a good address management
DHCPv6 is nearly useless (and implementors will restrict it to the
"other parameters" part as some have already done :-).

   >> 3.  Client sends Request with address extension or extensions
   >>     which are not empty.

   >Jim, I don't think this is a good scenario. A Client should never
   >be involved in telling a server about IP addresses (or other
   >releasable resources) it may have that the server has no knowledge
   >of. Sure, a client MAY request that a certain address be assigned to
   >it, and if the request is valid from a network topology point of view
   >and a administrator policy view, then fine. I just don't think we
   >should be configuring DHCP servers with DHCP clients, IMHO.
   
=> this is a matter of taste. For primary addresses I agree that the
server should have the control and the client should be kept as stupid/
simple as possible. For secondary addresses the context is very different
because the client is supposed to know what it wants and to use the
server only as a check/register device is fine for me.

   Same difference.  If the address is not topologically correct then the
   server would not permit it?
   
=> I think your idea is to restrict the server to a passive role too
in this case.

Thanks

Francis.Dupont@inria.fr


From dhcp-v6@bucknell.edu  Thu Dec  9 07:11:33 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id HAA29556
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 07:11:32 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id HAA27528;
	Thu, 9 Dec 1999 07:06:58 -0500 (EST)
Date: Thu, 9 Dec 1999 07:06:58 -0500 (EST)
Message-Id: <199912091202.NAA25926@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DISCUSSION: Use of Address Extensions with DHCPv6 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   I am getting a feeling that things are seeming much more complex
   than they really are.  I'll try to bring as much simplicity back
   into this conversation as I can possibly do.
   
=> do you think we should disagree? (:-)

   Allowing extensions to the advertisement message is a suitable way
   to allow for administrators to cater to sophisticated clients.  But
   on the other hand I don't think it means that all servers need to
   advertise everything they might be willing to offer to the clients,
   especially because so many clients won't care very much.
   
=> Charles' idea is just to offer addresses in advertisement as
the first part of the negociation for primary addresses.
We need something like this because the knowledge (of the prefix list)
is in the server and requests are done by clients.
This is only a (re)use of an existing mechanism in order to solve
the primary address management problem.

   > > 1.  Client sends DHCPv6 Request with no address extension.

=> this has been overruled by Charles' idea (if we agree about it).

   We decided in Minneapolis that all releaseable resources needed
   to be explicitly requested, even if right now everybody knew that IP
   address extensions could be designed otherwise.
   
=> I believe this is a good principle and if we can add a one for one
rule (ie. don't get 3 instances of a resource when only one was requested)
it will be better too.

   I think a client should be able to request an IP address even if the
   site is using stateless address autoconfiguration.  This does not lead
   to any ambiguities to my best current understanding.
   
=> if we adopt something like my primary/secondary classification then
in this case it will be a secondary IP address (because primaries were
dealt with by stateless autoconf).

   Furthermore, I don't think that the client has to request an address
   that is listed in an advertisement, and I don't think that the
   advertisement has to have an IP address extension before the client
   is allowed to get an IP address.
   
=> how do you solve the multiple prefix issue?
   
   >    At DC, I proposed a change to the IP address extension which would
   > permit the client to identify which network it wants
   > ...
   
   I don't see the need for this.

=> I agree
   
   If a client sends a nonempty IP address extension to a server, it
   is asking that server to renew or extend the lifetime for that address.

=> there is a third case: the secondary/extra address. And of course
the discussion is about that...
   
Thanks

Francis.Dupont@inria.fr


From dhcp-v6@bucknell.edu  Thu Dec  9 09:26:08 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA11457
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 09:26:08 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id JAA18765;
	Thu, 9 Dec 1999 09:17:33 -0500 (EST)
Date: Thu, 9 Dec 1999 09:17:33 -0500 (EST)
Message-Id: <199912091340.IAA08188@hygro.adsl.duke.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Thomas Narten <narten@raleigh.ibm.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: general releasable resources? 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

"HIBBS, BARR (PB)" <RBHIBBS@msg.pacbell.com> writes:

> However, I think there is one issue we must resolve first:  why must a
> client request a releasable resource, but not be required to explicitly
> request non-releasable ones?  If some extensions are intended to always be
> supplied by a server, are these specified in one of the drafts?  I actually
> favor a scheme where clients explicitly request all extensions that they
> desire, with the server supplying any that are required because of [other]
> protocol requirements or local site policy.  I just want to see the language
> about what is and which ones are automatically provided and which ones must
> be requested to be unambiguous in the RFCs.

My understanding of the reason for requiring that client's explicitly
request releasable resources is to avoid the problem in DHCPv4 where
it is problematical for the server to send options that the client
doesn't understand.

I too share your concern that one of the benefits of DHCP is allowing
the sys admin to say (via a configuration in the server) here is what
I want the client to have, independent of what the client actually
requests. So, maybe what is needed is not that the client explicitely
request a releasable resource, but a way for the client to indicate
that it supports a particular resource, should the server choose to
send it. But even that may be unnecessarily restrictive.

Thomas


From dhcp-v6@bucknell.edu  Thu Dec  9 09:27:30 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id JAA12207
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 09:27:29 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id JAA02634;
	Thu, 9 Dec 1999 09:15:58 -0500 (EST)
Date: Thu, 9 Dec 1999 09:15:58 -0500 (EST)
Message-Id: <199912091324.IAA08095@hygro.adsl.duke.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Thomas Narten <narten@raleigh.ibm.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

Charles Perkins <charliep@iprg.nokia.com> writes:

> 1. I can imagine a _lot_ of releasable resources in the future, but
>    it would be a mistake to try to specify them before we can have
>    DHCPv6 as a Proposed Standard.  After that time, new extensions
>    to specify new releasable resources will come up for discussion
>    in a natural way, as long as we have a natural mechanism specified
>    for handling them.  Some possible examples of releasable
>    resources:

>    = disk space
>    = network memory
>    = I/O devices
>    = Licensed software
>    = time-metered equipment
>    = ports
>    = protocol translators

After thinking about these examples a bit, I am not sure they are
compelling examples of a potential releasable resources. In each of
the above, they are not a resource like addresses are today in sense
that they are given to the client and the client then uses
them. Rather, they seem more like services/servers, in which a special
protocol is needed to actually use the service. For example, to use
remote disk space or remote memory, the client would need to use a
special protocol to access the server. Moreover, the protocol would
likely be specific to the service being provided. Thus, providing the
address of the server would be something that DHCP could do (or,
perhaps even better, SLP should be used to find it), but that's about
it. The client would use the service-specific protocol to actually
access the resource and negotiate with the server about its usage. So
I really don't see how these make sense as a releasable resource.

The licensed software might be an exception, in the sense that perhap
the software could be sent to the client and the client could then use
it for the duration of the lease. But even here, I suspect there would
be much more to providing such a solution than just having a
releasable resource via DHCP. For example, the code would probably be
large enough that  one wouldn't want to use DHCP to actually transfer
it, but that then means we have a special (non-DHCP) protocol and it
becomes unclear what benefit the DHCP portion provides.

Thomas


From dhcp-v4@bucknell.edu  Thu Dec  9 10:20:32 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA10162
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 10:20:32 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA06529;
	Thu, 9 Dec 1999 10:20:02 -0500 (EST)
Date: Thu, 9 Dec 1999 10:20:02 -0500 (EST)
Message-Id: <4.2.0.58.19991209092903.0162bac0@funnel.cisco.com>
Errors-To: droms@bucknell.edu
Reply-To: kkinnear@cisco.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Kim Kinnear <kkinnear@cisco.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 

At 08:40 PM 12/8/99 -0500, Greg Rabil wrote:
> > > I guess what I was trying to say was that since we decided that it
> > > was appropriate to put the load balancing functionality in another
> > > draft, it was less of a candidate for driving protocol
> > > details. Sounds like that would not meet with general agreement.
> >
> > Nope.   We just moved the technical description of load balancing out
> > of the draft.   It's still required by the draft, unless I am much
> > mistaken.   So it's as important as it was before the decision to
> > remove the details from the failover draft was made.
> >
> >                              _MelloN_
> >
>
>OK, now I'm confused.  I thought (like Richard) that we'd decided to move
>the load-balancing into a separate draft because is was _not_ required for
>failover.  Yes, it is nice to have, but I did not realize that we were
>requiring it.  If we are, then why publish a separate draft?  Is the
>failover draft now dependent upon the load-balancing draft for standard?
>
>Greg
>


I agree with Ted.  Load balancing is still required for a
complete failover implementation as specified in the -05 draft.
As I understood it, the reason to move the details of the
algorithm out of the failover draft and into a separate draft was
because the load balancing algorithm had applicability beyond
that of failover.

For instance, it would make sense if you just wanted to balance
the load between two regular servers, without any failover
involved.  Certainly there are situations where having two
servers handling the load would be cheaper than buying the
hardware to support one big server handling the same load.
Obviously you have to split your address pools, and this is a
challenge of sorts.  Of course, today people do this by putting
some pools on one server, and some on another, and get largely
the same effect, so that load balancing I think is really
interesting in a non-failover case mostly when you have *really*
big pools with *lots* of activity -- but some cases like this do
exist.

Cheers -- Kim





From dhcp-v4@bucknell.edu  Thu Dec  9 10:26:59 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA13577
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 10:26:58 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA15661;
	Thu, 9 Dec 1999 10:26:26 -0500 (EST)
Date: Thu, 9 Dec 1999 10:26:26 -0500 (EST)
Message-Id: <4.2.0.58.19991209093514.01618e40@funnel.cisco.com>
Errors-To: droms@bucknell.edu
Reply-To: kkinnear@cisco.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Kim Kinnear <kkinnear@cisco.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 


Richard said:

I don't think POOLREQ functions need to be expensive, if the
primary is implemented to know how many addresses the secondary
has and how many it's allowed (or configured) to have.  Maybe
that's going to be harder to do than I think ;-).

then Steve replied:

I am afraid it is a bit harder.  New trends in server operation
introduce complexities like arbitrary lease pools, (with
different settings and privileges), which leads to the primary
having to leaf through all of them, to hand over the prescribed
percantage of each.

Kim adds:

In my experience it depends entirely on how you implement the
accounting for lease pools.  If you keep up the number of primary
free and secondary free, then this is a very fast operation.  If
you don't, and have to calculate the numbers "on the fly", then
it isn't a very fast operation because you have to touch each
lease to determine if it is free and if so, who has it.

I would assert that you want to do it this first way in any case,
however, whether you do it on primary or secondary, because no
matter what, you don't want to have to scan all of the leases to
decide whether you need more or your partner needs more.

-----------------------------------------

Richard said:

I always assumed that POOLREQ/POOLRESP exchanges could happen any
time, and that when they did happen, there would be a period of
extra bind updates and then it would be done for a while.  But
compared to the rest of the traffic, it doesn't seem that awful,
especially since the folks who have implemented already seem to
feel that wire traffic is nothing compared to other functions
(like stable storage).

Steve replied:

It all depends on the kinds of trafic you are handling.  If you
want carrier class performance, wire traffic does become an
issue.

Kim adds:

I agree with Richard -- POOLREQ can happen an time.  From the
efficiency standpont I don't see any difference in the wire
traffic with either approach.  The actual POOLREQ is totally
insignificant in the greater scheme of things.  There is a
BNDUPD/BNDACK operation in either approach, from one server to
the other, here, so I don't find this particular argument to be
significant.

----------------------------------

Richard said:

If we have the secondary change the state of a lease to BACKUP
and update the primary, what happens in the case of a conflict?
Maybe we'd simply need another reject-reason code, but there may
be more to it than that.

Steve replied:

I would very much like to avoid cycling the backup server's expiring
leases through the primary if it can be avoided, for reasons of
efficiency, and keeping the same lease with the same server.

If someone can come up with a good reason why we have to do this,
I would very much like to hear it.

Kim adds:

I don't think there is an efficiency issue here, frankly.  As I
noted above, there is a BNDUPD/BNDACK in any case before the
backup can use the IP address for a different client.  Whether
the BNDUPD/BNDACK is initiated by the primary or secondary is the
issue, I think.

I need to think about this for a bit, since something bothers me
about the approach of lease ownership of leased leases.  The
basic design for this was done a number of years ago, actually,
and so it may take a bit to recover some of the thinking behind
the current approach.  

In addition, I'm moving my office today and we have a release
about to ship just now, so I'm going to be out of touch for at
least a few days I expect.  But I will continue to think about
this issue and try to figure out what concerns me about it.

Cheers - Kim

  


From dhcp-v4@bucknell.edu  Thu Dec  9 10:31:52 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA16263
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 10:31:51 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA31439;
	Thu, 9 Dec 1999 10:31:34 -0500 (EST)
Date: Thu, 9 Dec 1999 10:31:34 -0500 (EST)
Message-Id: <NCBBLHHPJKGMBMMHDNMMIEOLCNAA.derdmann@nortelnetworks.com>
Errors-To: droms@bucknell.edu
Reply-To: derdmann@nortelnetworks.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: "David Erdmann" <derdmann@nortelnetworks.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2910.0)
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset="iso-8859-1"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit

My understanding was that if there was any dependency that the load
balancing was dependent on the failover draft.  But that the failover
draft could be implemented without the load balancing.

I agree that requirements for the load balancing should play an important
role in defining the failover draft, but I dont' think that we should
require load balancing in order to implement the failover.

-Dave

>
> > > I guess what I was trying to say was that since we decided that it
> > > was appropriate to put the load balancing functionality in another
> > > draft, it was less of a candidate for driving protocol
> > > details. Sounds like that would not meet with general agreement.
> >
> > Nope.   We just moved the technical description of load balancing out
> > of the draft.   It's still required by the draft, unless I am much
> > mistaken.   So it's as important as it was before the decision to
> > remove the details from the failover draft was made.
> >
> > 			       _MelloN_
> >
>
> OK, now I'm confused.  I thought (like Richard) that we'd decided to move
> the load-balancing into a separate draft because is was _not_ required for
> failover.  Yes, it is nice to have, but I did not realize that we were
> requiring it.  If we are, then why publish a separate draft?  Is the
> failover draft now dependent upon the load-balancing draft for standard?
>
> Greg



From dhcp-v6@bucknell.edu  Thu Dec  9 11:45:17 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA24617
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 11:45:16 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA26831;
	Thu, 9 Dec 1999 11:41:51 -0500 (EST)
Date: Thu, 9 Dec 1999 11:41:51 -0500 (EST)
Message-Id: <384FD17C.BE40DDF6@iprg.nokia.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Charles E. Perkins" <charliep@iprg.nokia.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: general releasable resources?
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
Content-Transfer-Encoding: 7bit

Thomas Narten wrote:


> I too share your concern that one of the benefits of DHCP is allowing
> the sys admin to say (via a configuration in the server) here is what
> I want the client to have, independent of what the client actually
> requests. So, maybe what is needed is not that the client explicitely
> request a releasable resource, but a way for the client to indicate
> that it supports a particular resource, should the server choose to
> send it. But even that may be unnecessarily restrictive.

I think this is a good formulation.  Assuming we have an
"extension request extension", it would be simple for a client
to include the resources it can support receiving from the server.
If the server is not configured to send the resource, the client
cannot force it to do so.

Regards,
Charlie P.


From dhcp-v6@bucknell.edu  Thu Dec  9 11:46:39 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA25361
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 11:46:36 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA10450;
	Thu, 9 Dec 1999 11:41:04 -0500 (EST)
Date: Thu, 9 Dec 1999 11:41:04 -0500 (EST)
Message-Id: <384FD0DE.266EE011@iprg.nokia.com>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "Charles E. Perkins" <charliep@iprg.nokia.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DC meeting: Discussion question #1, releaseable resources
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
Content-Transfer-Encoding: 7bit

Thomas Narten wrote:

>>    = disk space
>>    = network memory
>>    = I/O devices
>>    = Licensed software
>>    = time-metered equipment
>>    = ports
>>    = protocol translators
> 
> After thinking about these examples a bit, I am not sure they are
> compelling examples of a potential releasable resources. In each of
> the above, they are not a resource like addresses are today in sense
> that they are given to the client and the client then uses
> them. Rather, they seem more like services/servers, in which a special
> protocol is needed to actually use the service.

(>> == me)
(>  == Thomas)
I am wholly in favor of using SLP for resolving service needs, and
I picked those examples exactly because I don't think they need SLP.

>                                                For example, to use
> remote disk space or remote memory, the client would need to use a
> special protocol to access the server.

So, for instance, I would expect that there might be an NFS scratch
space resource extension.  This dependency on protocol is reflected
today in DHCPv4.  Furthermore, it is reflected even in other extensions
for non-releasable resources.  For instance, one asks for an NTP
timeserver, not just a timeserver.  And, one asks for a DNS server,
not just a name directory server.

>                   ......              Thus, providing the
> address of the server would be something that DHCP could do

Yes.

>                                                                 (or,
> perhaps even better, SLP should be used to find it),

Probably not in those cases.

>     The client would use the service-specific protocol to actually
> access the resource and negotiate with the server about its usage. So
> I really don't see how these make sense as a releasable resource.

It makes sense for the DHCP server to identify servers for the client,
when the DHCP server has been configured to do so.  It makes sense for
disk space per subnet, network memory per subnet, and I think in every
one of the other cases that I enumerated.

> The licensed software might be an exception, in the sense that perhap
> the software could be sent to the client and the client could then use
> it for the duration of the lease. But even here, I suspect there would
> be much more to providing such a solution than just having a
> releasable resource via DHCP. For example, the code would probably be
> large enough that  one wouldn't want to use DHCP to actually transfer
> it, but that then means we have a special (non-DHCP) protocol and it
> becomes unclear what benefit the DHCP portion provides

To be clear, I do not say that we have designed the protocols for each
of the other cases.  I also do not say that anyone has designed the
extension.  I only offer them as examples that are somewhat realistic.
Until the details are worked out, it is probably all to easy to find out
things that _might_ be wrong with each of them.

It could also be the case that there would be competing ideas.  Some
people might want to model disk space as a SLP service.  It's an idea
that is worth exploring.  My feeling is that the only service attributes
that matter are location and size of the space requirement, and that is
not so cogent an argument for use for SLP vs. DHCP.  But, other things
might matter - like latency, transfer speed, and so on.  Furthermore,
since I lost the "protocol" argument in SLP, it is a moot point whether
"protocol" would be a service attribute for selection by way of SLP.
Thus, this is not a distinguishing characteristic for determining
whether to use SLP or DHCP.

For the purposes of this discussion, we are only trying to figure out
whether to worry about specifying behavior now for releasable resources.
My vote is to delay the actual layout of a general packet format for
such resources, and just try to make sure that DHCP _could_ support
such things in the future.  I think we are well along the way to making
that happen.

Regards,
Charlie P.


From dhcp-v6@bucknell.edu  Thu Dec  9 11:49:16 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA26718
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 11:49:16 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA24368;
	Thu, 9 Dec 1999 11:42:05 -0500 (EST)
Date: Thu, 9 Dec 1999 11:42:05 -0500 (EST)
Message-Id: <199912091628.LAA13836@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: WG: general releasable resources?
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 04:38 PM 12/8/99 -0500, you wrote:
>Do we attempt to specify general releasable resources now, or should we
>defer their specification till later if we find them useful?

The work we've done in defining releasable resources has convinced me
that the cost of adding releasable resources to the protocol spec (including
additional work to complete the definition, add text to the spec and
confirm the correctness of the spec) and
implementation outweighs the potential benefit - especially
as the only example of a releasable resource we've come up with
to date has been IPv6 addresses.

I would suggest we leave hooks for future definition of releasable
resources in the spec and defer additional work on releasable resources
until we have a more compelling case in support of them.

- Ralph




From dhcp-v6@bucknell.edu  Thu Dec  9 13:02:26 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA05553
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 13:02:25 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA14663;
	Thu, 9 Dec 1999 12:52:19 -0500 (EST)
Date: Thu, 9 Dec 1999 12:52:19 -0500 (EST)
Message-Id: <199912091715.SAA11405@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: WG: general releasable resources? 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   The work we've done in defining releasable resources has convinced me
   that the cost of adding releasable resources to the protocol spec (including
   additional work to complete the definition, add text to the spec and
   confirm the correctness of the spec) and
   implementation outweighs the potential benefit - especially
   as the only example of a releasable resource we've come up with
   to date has been IPv6 addresses.
   
=> I believe a part of your statement is missing (and you think this cost
is too high).

   I would suggest we leave hooks for future definition of releasable
   resources in the spec and defer additional work on releasable resources
   until we have a more compelling case in support of them.
   
=> I agree: I prefer a working DHCPv6 spec ASAP.

Thanks

Francis.Dupont@inria.fr


From dhcp-v6@bucknell.edu  Thu Dec  9 13:58:05 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA04584
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 13:58:03 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA03921;
	Thu, 9 Dec 1999 13:44:58 -0500 (EST)
Date: Thu, 9 Dec 1999 13:44:58 -0500 (EST)
Message-Id: <199912091806.NAA17051@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: WG: general releasable resources? 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

At 12:53 PM 12/9/99 -0500, Francis Dupont wrote:
> In your previous mail you wrote:
>
>   The work we've done in defining releasable resources has convinced me
>   that the cost of adding releasable resources to the protocol spec (including
>   additional work to complete the definition, add text to the spec and
>   confirm the correctness of the spec) and
>   implementation outweighs the potential benefit - especially
>   as the only example of a releasable resource we've come up with
>   to date has been IPv6 addresses.
>   
>=> I believe a part of your statement is missing (and you think this cost
>is too high).

I meant to imply that the cost is "too high" with the phrase "outweighs the
potential benefit"...

>   I would suggest we leave hooks for future definition of releasable
>   resources in the spec and defer additional work on releasable resources
>   until we have a more compelling case in support of them.
>   
>=> I agree: I prefer a working DHCPv6 spec ASAP.
>
>Francis.Dupont@inria.fr

- Ralph




From dhcp-v4@bucknell.edu  Thu Dec  9 16:09:26 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA13012
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 16:09:26 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA06430;
	Thu, 9 Dec 1999 16:08:41 -0500 (EST)
Date: Thu, 9 Dec 1999 16:08:41 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE7009@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0


>In my experience it depends entirely on how you implement the
>accounting for lease pools.  If you keep up the number of primary
>free and secondary free, then this is a very fast operation.  If
>you don't, and have to calculate the numbers "on the fly", then
>it isn't a very fast operation because you have to touch each
>lease to determine if it is free and if so, who has it.

OK. This is certainly a good optimization. 

Let's look at the message exchange you need when you start with a 
lease in backup state, to put that lease into backup state again.

1) send a binding update, inticating the lease is active.
2) receive ack
3) send a binding update indicating the lease is now expired/released
4) reveive ack
--- at this point you can not use the lease because it is in "free" state,
not "backup".--
5) send a poolrequest
6) receive a poolreply
7) receive a binding update indicating the lease is now "backup".
8) send an ack.

I am arguing that 5)..8) is not necessary, if you put the lease back in
"backup" state
in 4).


>I agree with Richard -- POOLREQ can happen an time.  From the
>efficiency standpont I don't see any difference in the wire
>traffic with either approach.  The actual POOLREQ is totally
>insignificant in the greater scheme of things. 

No argument there. It is the extra update traffic
needed to bounce back the expired leases that bothers me.

>I don't think there is an efficiency issue here, frankly.  As I
>noted above, there is a BNDUPD/BNDACK in any case before the
>backup can use the IP address for a different client. 

No, it can not use the address for a new client, 
because it no longer owns it!

sG


From dhcp-v6@bucknell.edu  Thu Dec  9 17:11:15 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA12958
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 17:11:14 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id RAA20144;
	Thu, 9 Dec 1999 17:05:02 -0500 (EST)
Date: Thu, 9 Dec 1999 17:05:02 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.944776852.25628.nordmark@jurassic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Erik Nordmark <Erik.Nordmark@eng.sun.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 Input from Francis Dupont 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0

> => there are two different problems:
>  - how the server identifies the client (and should the server recognize
>    a client using a different relay?)
>  - how the relay gets the link to use in a packet from a server.
> My concern is for the second problem the only good solution is to have
> the interface index because a naked address is not enough.

Agreed that ifindex would work for the 2nd problem. 
I thought you were talking about the 1st problem.

   Erik



From dhcp-v4@bucknell.edu  Thu Dec  9 18:21:00 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA18278
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 18:21:00 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id SAA22938;
	Thu, 9 Dec 1999 18:20:32 -0500 (EST)
Date: Thu, 9 Dec 1999 18:20:32 -0500 (EST)
Message-Id: <385036D3.D30D91C0@metaip.checkpoint.com>
Errors-To: droms@bucknell.edu
Reply-To: richardj@metaip.checkpoint.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Richard Jones <richardj@metaip.checkpoint.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
Content-Transfer-Encoding: 7bit



Steve Gonczi wrote:

> Let's look at the message exchange you need when you start with a
> lease in backup state, to put that lease into backup state again.
>
> 1) send a binding update, inticating the lease is active.
> 2) receive ack
> 3) send a binding update indicating the lease is now expired/released
> 4) reveive ack
> --- at this point you can not use the lease because it is in "free" state,
> not "backup".--
> 5) send a poolrequest
> 6) receive a poolreply
> 7) receive a binding update indicating the lease is now "backup".
> 8) send an ack.
>
> I am arguing that 5)..8) is not necessary, if you put the lease back in
> "backup" state
> in 4).

Having had another day to think about this, I find another question.
One of the design goals in RFC 2131 is to 'Retain DHCP client
configuration across DHCP client reboot. A DHCP client should ...be
assigned the same configuration parameters (e.g., network address)
in response to each request.'

It seems like having to cycle leases through the primary each time
their state changes away from active makes this very difficult.Any
comments on this? Am I drifting over to Steve's point of view?
Don't know yet - the more I think...

Thanks, Steve. This gets my mind off shipping dates!

Regards,
Richard



From dhcp-v4@bucknell.edu  Thu Dec  9 19:02:15 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA08325
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 19:02:13 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA25219;
	Thu, 9 Dec 1999 19:00:50 -0500 (EST)
Date: Thu, 9 Dec 1999 19:00:50 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE700A@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0

>It seems like having to cycle leases through the primary each time
>their state changes away from active makes this very difficult.Any
>comments on this? Am I drifting over to Steve's point of view?
>Don't know yet - the more I think...

I do not think there is a problem with this, because the 2 servers
hopefully replicate all information relevant to clients.

Stay tuned, I am composing another message, where I am voicing some 
more dissenting views. ;-)

sG


From dhcp-v4@bucknell.edu  Thu Dec  9 19:03:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA08933
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 19:03:39 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA21315;
	Thu, 9 Dec 1999 19:03:15 -0500 (EST)
Date: Thu, 9 Dec 1999 19:03:15 -0500 (EST)
Message-Id: <716D440F8C29D311991100A0C920487423D973@GIDGET>
Errors-To: droms@bucknell.edu
Reply-To: Andre@incognito.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: "Kostur, Andre" <Andre@incognito.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Configuration Management
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0

I'm curious as to how the rest of the implementors are approaching the
problem of configuration management between the primary and secondary
servers.  As far as I can tell, the draft just says (implies?) that the two
servers are configured identically, and that's that.  However one must still
deal with the problem.  Does a configuration change mean that both servers
have to be effectively stopped, apply the configuration change, then restart
the services?  Or do the two servers talk to each other about config changes
when they happen?  And in this case, what if there are config changes during
COMMUNICATIONS-INTERRUPTED or PARTNER-DOWN states (ok, basically any state
other than NORMAL....)?


From dhcp-v4@bucknell.edu  Thu Dec  9 19:05:34 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA09762
	for <dhc-archive@odin.ietf.org>; Thu, 9 Dec 1999 19:05:33 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA05430;
	Thu, 9 Dec 1999 19:05:11 -0500 (EST)
Date: Thu, 9 Dec 1999 19:05:11 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE700B@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0

Hello,

I pondered a bit more on the POOLREQUEST & related issues.
I respectfully put forward, that the current lease status 
design is flawed. The error we have made is reminescent of
the Y2K glitch.

Leases are going through the EXACT SAME changes on both the
primary, and the secondary. (Get assigned, expire, etc).
These identical changes, and nothing else, should be reflected 
in the lease status. The lease ownership is an entirely 
different kind of thing. Hence:::

The lease ownership should be a separate attribute, 
(handled independently of the lease status).

If it was so, leases ownership could be transferred 
even if the lease was in active state, or if it was a
reservation (BTW we need a new lease state for 
reservations). 

IMHO this would make it a lot easier to conceptualize 
things:

a) Leases belong to either the primary or the secondary.
This is decided by a POOLREQUEST/REPLY negotiation.
(The Binging Update message needs a 
new option for lease ownership. )

b) The ownership assignment remains in effect, until 
a renegotiation takes place. The sole purpose of a renegotiation
is to reflect configuration changes ( e.g: new lease pools
are added, or the backup pool percentage has been changed).

c) Negotiation takes place after synchronization is complete.

d) The backup server does not start serving clients until 
the address ownership negotiation is done.

e) Leases can be given, or taken away by the primary server
at its discretion, even if they are being used by a client.
This is no big deal, as either server is allowed to renew
a lease. It also makes it easy to set the exact backup/primary
lease percentages.

I have a bit more to say, but let's see how this goes over ;-).

sG  


From dhcp-v4@bucknell.edu  Fri Dec 10 10:33:29 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA25181
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 10:33:28 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA28164;
	Fri, 10 Dec 1999 10:33:06 -0500 (EST)
Date: Fri, 10 Dec 1999 10:33:06 -0500 (EST)
Message-Id: <199912101455.JAA00343@uplink.net>
Errors-To: droms@bucknell.edu
Reply-To: droms@bucknell.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Web-based mailing list archive
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

Please check out this prototype web archive for the dhcp-v4 mailing list
and send me any feedback you might have:


http://www.listproc.bucknell.edu/cgi-bin/archdex?list=dhcp-v4

In production, the archive would be updated every hour so the latest
messages would be available...

- Ralph




From dhcp-v6@bucknell.edu  Fri Dec 10 10:34:44 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA25230
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 10:34:44 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA06012;
	Fri, 10 Dec 1999 10:28:58 -0500 (EST)
Date: Fri, 10 Dec 1999 10:28:58 -0500 (EST)
Message-Id: <199912101455.JAA00382@uplink.net>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Web-based mailing list archive
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

Please check out this prototype web archive for the dhcp-v6 mailing list
and send me any feedback you might have:


http://www.listproc.bucknell.edu/cgi-bin/archdex?list=dhcp-v6

In production, the archive would be updated every hour so the latest
messages would be available...

- Ralph




From dhcp-v4@bucknell.edu  Fri Dec 10 12:45:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28807
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 12:45:40 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA19518;
	Fri, 10 Dec 1999 12:44:31 -0500 (EST)
Date: Fri, 10 Dec 1999 12:44:31 -0500 (EST)
Message-Id: <199912101735.MAA08974@uplink.net>
Errors-To: droms@bucknell.edu
Reply-To: droms@bucknell.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Minutes from DC WG meetings
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

Here are the draft minutes from the DC WG meetings.  Please
review and post comments by this afternoon as the minutes are
due today.

I'm sorry for the short notice for review and comment on
the minutes...

- Ralph

Tuesday, Nov 9 at 0900-1130 (DHCPv4)
====================================

Dynamic host configuration : DHCP reconfigure extension Yves T'Joens (Alcatel)
draft-schrijvp-dhcpv4-reconfigure-00.txt

The author described a proposed option for a unicast message to
trigger DHCP client reconfiguration.  The WG provided feedback (add
authentication, make draft more precise) and agreed to consider the
option after a new draft is submitted.


The Name Service Search Option for DHCP                 Carl Smith (Sun)
draft-ietf-dhc-nsso-01.txt

The WG suggested using a 16-bit value for the service numbers, and to
define 0 as a reserved value to configure the client to look in local
configuration table.  The draft will be ready for WG last call after a
final small revision.


DHCP Failover Protocol                                  Kim Kinnear (Cisco)
draft-ietf-dhc-failover-05.txt

Kim reported on phone conferences held since last WG meeting.  He
will schedule one or two additional phone conferences and submit a new
draft for (well before?) the next IETF.


The User Class Option for DHCP                          Jerome Privat (BT)
draft-ietf-dhc-userclass-04.txt

The WG provided feedback on the most recent draft: simplify the draft,
reduce "MUST" policy directives.  The author will revise draft
according to WG feedback.


Double Phase DHCP Configuration                         Jerome Privat (BT)
draft-privat-dhc-doublephase-01.txt

WG suggested splitting this document into two drafts: an informational
draft for the proxy mechanism and a standards track draft for redirect
option.  WG will continue to review this option after new drafts are
published. 


Dynamic Registration and Configuration Protocol (DRCP)  Anthony McAuley
draft-itsumo-drcp-00.txt                                (Telcordia)

Protocol related to DHCP that reduces DHCP client-server communication
latency through distributed DHCP servers (or proxies).  The WG
expressed the opinion that this may be a new protocol, not DHCP.  The
WG agreed to look at the next draft and decide whether or not to
review the protocol.


Authentication for DHCP Messages                        Bill Arbaugh (Penn)
draft-ietf-dhc-authentication-12.txt

Bill reported on suggestions from Volz. Ran Atkinson suggested
splitting into two drafts; WG chose to leave "as is".  Draft will go
to WG last call after next revision.


DHCP Authentication Via Kerberos V                      Bernard Aboba
draft-hornstein-dhc-kerbauth-01.txt                     (Microsoft)

Bernard presented an extension to DHCP to provide Kerberos V
authentication between clients and servers and to provide interdomain
authentication.  The basic mechanism is to transport Kerberos
authentication data in DHCP messages (with, potentially, other
communication between Kerberos participants).  The WG suggested
external, authoritative review of the application of Kerberos
technology and took no specific action.


DHC load balancing algorithm                            Bernie Volz
draft-ietf-dhc-loadb-00.txt                             (Process Software)

This draft results from load balancing technology developed with
failover protocol, which seemed to have wider application to DHCP
server configuration.  Mike Patrick expressed concern about the

necessity of configuring relay agents to match servers (if not, some
requests may get dropped).  Bernie will revise draft according to
feedback from WG, and WG will continue to review the document.


The Server Selection Option for DHCP                    Bill Sommerfield
draft-ietf-dhc-sso-03.txt                               (Epilogue)

Rob Stevens suggested this option overlaps with load balancing; WG
consensus was that overlap was small enough to warrant continued
consideration of the server selection option.  The WG suggested
simplifying the draft (e.g., the descriptions of profiles), changing
several "MUST"s to "SHOULD"s, and to allow lease time as tie breaker.
Bill will revise and the WG will review the next draft.


Dynamic Host Configuration Protocol (DHCP) Server MIB   Barr Hibbs (Pacbell)
draft-ietf-dhc-server-mib-04.txt

Barr will post specific questions to the WG mailing list and revise
MIB according to feedback.  Next revision should be ready for WG last
call. 


Interpreting Client Options for DHCP                    Barr Hibbs (Pacbell)
draft-ietf-dhc-client-options-00.txt

Barr will withdraw this draft or let it expire.  Ted Lemon volunteered
to begin work on a revision of RFC 2131/2132 that will be submitted
for full Standard status.  This revision will include clarifications
and revised text based on Barr's draft as well as other experience
with DHCP.


Option for NAT/IPSec interaction                        Jose Brustoloni
(no draft available)                                    (Bell Labs)

This option addresses interoperation problems between NAT and IPSec.
WG provided feedback and will review a revised draft.


Subnet selection option                                 Ted Lemon
draft-ietf-dhc-subnet-option-03.txt                     (for Glenn Waters)

Draft is ready for WG last call.


Static routes with subnet masks option                  Ted Lemon
(no draft available)

Ted will write up draft for an option to pass a list of static routes
with subnet masks.


Domain name search list option                          Ted Lemon
draft-ietf-dhc-domsrch-02.txt


Ted will track down previous draft and move it forward.


Tuesday, Nov 9 at 1300-1400 (DHCPv4)
====================================

DHCP Relay Agent Information Option                     Mike Patrick (Motorola)
draft-ietf-dhc-agent-options-07.txt

Mike described Motorola IPR statement.  A few new issues were raised:
subnet sub-option (WG consensus was to keep it); interaction between
relay agent option and IPSec (relay agent should not add option);
server should not assume relay agent is present and may choose to add
a relay agent information option.  The WG agreed to hold a last call
on a revised document that addresses these points.


DHCP Schema for LDAP                                    Bernie Volz
draft-ietf-dhc-schema-01.txt                            (Process Software)

Changes since last revision:

* New class, containment and inheritance hierarchy
* Uses common name attribute
* Added dhcpPolicy object class and text on mapping dhcpPolicy object
  to Policy WG schema
* Removed use of auxiliary classes

WG agreed to move forward as Proposed Standard document with
(potentially?) last call after next revision.


Interaction between DHCP and DNS                        Mark Stapp (Cisco)
draft-ietf-dhc-dhcp-dns-11.txt

A new RR for use with DHCP has been proposed.  DHCP will use this
instead of the previously proposed use of the KEY RR.  Also, the
security section must be updated to reflect TSIG.  Mark will
revise and resubmit the draft to reflect these changes.


A DNS RR for DHCP information                           Andreas Gustafsson
draft-ietf-dnsind-dhcp-rr-00.txt                        (Internet Engines)

The DHCP RR (name will likely change) is an opaque data item stored by
a DNS server and holds a DHCP client identifier to indicate the client
currently associated with a DNS entry.  The new RR holds only a client
identifier and no other information (e.g., lease).  The WG consensus
was that the DHCP RR will meet the requirements for DHCP-DNS
interaction and should be moved forward.


New Option Review Guidelines and                        Mike Carney (Sun)
  Additional Option Namespace
draft-ietf-dhc-option-review-and-namespace-01.txt

Mike solicited comments from the WG on section 5 (open issues) and
would like to go to last call after minor revisions addressing those
issues.


Tuesday, Nov 9 at 1415-1515 (DHCPv6)
====================================
Dynamic Host Configuration Protocol for IPv6 (DHCPv6)   Mike Carney (Sun)
draft-ietf-dhc-dhcpv6-14.txt                            Jim  Bound (Compaq)
Extensions for the DHCP for IPv6                        Charlie Perkins (Nokia)
draft-ietf-dhc-v6exts-11.txt

WG discussed several open issues:

* Releasable resources: IPv6 addresses are one example; are there
  others (none suggested).  When and how should these resources be
  managed (what is leasing strategy) and released (to original server
  or any server)?
* Stateful autoconfiguration and multiple IP addresses: discussion
  centered on tradeoff between complexity and utility - client might
  make multiple requests for addresses in IPv4-IPv6 transition.
  Discussion will continue on DHCPv6 mailing list.
* Stateful autoconfiguration and use of multiple servers (e.g.,
  failover): any useful input from experience with DHCPv4?
* Document organization: authors prefer to leave as two documents.

Authors will pose questions separately to DHCPv6 mailing list for WG
discussion and publish revised draft by end of calendar year.




From dhcp-v4@bucknell.edu  Fri Dec 10 12:48:25 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28876
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 12:48:23 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA23005;
	Fri, 10 Dec 1999 12:47:31 -0500 (EST)
Date: Fri, 10 Dec 1999 12:47:31 -0500 (EST)
Message-Id: <38513AF4.AE549EDC@metaip.checkpoint.com>
Errors-To: droms@bucknell.edu
Reply-To: richardj@metaip.checkpoint.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Richard Jones <richardj@metaip.checkpoint.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
Content-Transfer-Encoding: 7bit


Steve Gonczi wrote:

> >It seems like having to cycle leases through the primary each time
> >their state changes away from active makes this very difficult.Any
> >comments on this? Am I drifting over to Steve's point of view?
> >Don't know yet - the more I think...
>
> I do not think there is a problem with this, because the 2 servers
> hopefully replicate all information relevant to clients.

Hold on a sec. With our current setup, let's say the secondary is
configured via hash bucket assignment to respond to a DISCOVER
from client A. It assigns an IP, and sometime later client A releases
it. The address then gets cycled back to the primary. Some time after
this, client A does either another DISCOVER or a REQUEST, with
every right to expect the secondary to make all efforts to give it back
the same IP. However, since the address has been cycled back to
the primary, and the mechanism for getting addresses back to the
secondary is based more on whether or not the secondary is
running low on addresses, I think this design goal goes by the
wayside. Am I missing something?

I guess the only way to deal with this right now is for the secondary
to issue a POOLREQ every time it relinquishes an IP back to the
primary. I don't think this is what we had in mind with the
POOLREQ/POOLRESP messages. If I were to be real stubborn
about my earlier argument concerning the primary's priority, I'd
say this was okay, but now I'm wondering if it's necessary.

> Stay tuned, I am composing another message, where I am voicing some
> more dissenting views. ;-)
>
> sG



From dhcp-v6@bucknell.edu  Fri Dec 10 12:48:38 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28904
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 12:48:38 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA30497;
	Fri, 10 Dec 1999 12:40:44 -0500 (EST)
Date: Fri, 10 Dec 1999 12:40:44 -0500 (EST)
Message-Id: <199912101735.MAA08974@uplink.net>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Minutes from DC WG meetings
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

Here are the draft minutes from the DC WG meetings.  Please
review and post comments by this afternoon as the minutes are
due today.

I'm sorry for the short notice for review and comment on
the minutes...

- Ralph

Tuesday, Nov 9 at 0900-1130 (DHCPv4)
====================================

Dynamic host configuration : DHCP reconfigure extension Yves T'Joens (Alcatel)
draft-schrijvp-dhcpv4-reconfigure-00.txt

The author described a proposed option for a unicast message to
trigger DHCP client reconfiguration.  The WG provided feedback (add
authentication, make draft more precise) and agreed to consider the
option after a new draft is submitted.


The Name Service Search Option for DHCP                 Carl Smith (Sun)
draft-ietf-dhc-nsso-01.txt

The WG suggested using a 16-bit value for the service numbers, and to
define 0 as a reserved value to configure the client to look in local
configuration table.  The draft will be ready for WG last call after a
final small revision.


DHCP Failover Protocol                                  Kim Kinnear (Cisco)
draft-ietf-dhc-failover-05.txt

Kim reported on phone conferences held since last WG meeting.  He
will schedule one or two additional phone conferences and submit a new
draft for (well before?) the next IETF.


The User Class Option for DHCP                          Jerome Privat (BT)
draft-ietf-dhc-userclass-04.txt

The WG provided feedback on the most recent draft: simplify the draft,
reduce "MUST" policy directives.  The author will revise draft
according to WG feedback.


Double Phase DHCP Configuration                         Jerome Privat (BT)
draft-privat-dhc-doublephase-01.txt

WG suggested splitting this document into two drafts: an informational
draft for the proxy mechanism and a standards track draft for redirect
option.  WG will continue to review this option after new drafts are
published. 


Dynamic Registration and Configuration Protocol (DRCP)  Anthony McAuley
draft-itsumo-drcp-00.txt                                (Telcordia)

Protocol related to DHCP that reduces DHCP client-server communication
latency through distributed DHCP servers (or proxies).  The WG
expressed the opinion that this may be a new protocol, not DHCP.  The
WG agreed to look at the next draft and decide whether or not to
review the protocol.


Authentication for DHCP Messages                        Bill Arbaugh (Penn)
draft-ietf-dhc-authentication-12.txt

Bill reported on suggestions from Volz. Ran Atkinson suggested
splitting into two drafts; WG chose to leave "as is".  Draft will go
to WG last call after next revision.


DHCP Authentication Via Kerberos V                      Bernard Aboba
draft-hornstein-dhc-kerbauth-01.txt                     (Microsoft)

Bernard presented an extension to DHCP to provide Kerberos V
authentication between clients and servers and to provide interdomain
authentication.  The basic mechanism is to transport Kerberos
authentication data in DHCP messages (with, potentially, other
communication between Kerberos participants).  The WG suggested
external, authoritative review of the application of Kerberos
technology and took no specific action.


DHC load balancing algorithm                            Bernie Volz
draft-ietf-dhc-loadb-00.txt                             (Process Software)

This draft results from load balancing technology developed with
failover protocol, which seemed to have wider application to DHCP
server configuration.  Mike Patrick expressed concern about the

necessity of configuring relay agents to match servers (if not, some
requests may get dropped).  Bernie will revise draft according to
feedback from WG, and WG will continue to review the document.


The Server Selection Option for DHCP                    Bill Sommerfield
draft-ietf-dhc-sso-03.txt                               (Epilogue)

Rob Stevens suggested this option overlaps with load balancing; WG
consensus was that overlap was small enough to warrant continued
consideration of the server selection option.  The WG suggested
simplifying the draft (e.g., the descriptions of profiles), changing
several "MUST"s to "SHOULD"s, and to allow lease time as tie breaker.
Bill will revise and the WG will review the next draft.


Dynamic Host Configuration Protocol (DHCP) Server MIB   Barr Hibbs (Pacbell)
draft-ietf-dhc-server-mib-04.txt

Barr will post specific questions to the WG mailing list and revise
MIB according to feedback.  Next revision should be ready for WG last
call. 


Interpreting Client Options for DHCP                    Barr Hibbs (Pacbell)
draft-ietf-dhc-client-options-00.txt

Barr will withdraw this draft or let it expire.  Ted Lemon volunteered
to begin work on a revision of RFC 2131/2132 that will be submitted
for full Standard status.  This revision will include clarifications
and revised text based on Barr's draft as well as other experience
with DHCP.


Option for NAT/IPSec interaction                        Jose Brustoloni
(no draft available)                                    (Bell Labs)

This option addresses interoperation problems between NAT and IPSec.
WG provided feedback and will review a revised draft.


Subnet selection option                                 Ted Lemon
draft-ietf-dhc-subnet-option-03.txt                     (for Glenn Waters)

Draft is ready for WG last call.


Static routes with subnet masks option                  Ted Lemon
(no draft available)

Ted will write up draft for an option to pass a list of static routes
with subnet masks.


Domain name search list option                          Ted Lemon
draft-ietf-dhc-domsrch-02.txt


Ted will track down previous draft and move it forward.


Tuesday, Nov 9 at 1300-1400 (DHCPv4)
====================================

DHCP Relay Agent Information Option                     Mike Patrick (Motorola)
draft-ietf-dhc-agent-options-07.txt

Mike described Motorola IPR statement.  A few new issues were raised:
subnet sub-option (WG consensus was to keep it); interaction between
relay agent option and IPSec (relay agent should not add option);
server should not assume relay agent is present and may choose to add
a relay agent information option.  The WG agreed to hold a last call
on a revised document that addresses these points.


DHCP Schema for LDAP                                    Bernie Volz
draft-ietf-dhc-schema-01.txt                            (Process Software)

Changes since last revision:

* New class, containment and inheritance hierarchy
* Uses common name attribute
* Added dhcpPolicy object class and text on mapping dhcpPolicy object
  to Policy WG schema
* Removed use of auxiliary classes

WG agreed to move forward as Proposed Standard document with
(potentially?) last call after next revision.


Interaction between DHCP and DNS                        Mark Stapp (Cisco)
draft-ietf-dhc-dhcp-dns-11.txt

A new RR for use with DHCP has been proposed.  DHCP will use this
instead of the previously proposed use of the KEY RR.  Also, the
security section must be updated to reflect TSIG.  Mark will
revise and resubmit the draft to reflect these changes.


A DNS RR for DHCP information                           Andreas Gustafsson
draft-ietf-dnsind-dhcp-rr-00.txt                        (Internet Engines)

The DHCP RR (name will likely change) is an opaque data item stored by
a DNS server and holds a DHCP client identifier to indicate the client
currently associated with a DNS entry.  The new RR holds only a client
identifier and no other information (e.g., lease).  The WG consensus
was that the DHCP RR will meet the requirements for DHCP-DNS
interaction and should be moved forward.


New Option Review Guidelines and                        Mike Carney (Sun)
  Additional Option Namespace
draft-ietf-dhc-option-review-and-namespace-01.txt

Mike solicited comments from the WG on section 5 (open issues) and
would like to go to last call after minor revisions addressing those
issues.


Tuesday, Nov 9 at 1415-1515 (DHCPv6)
====================================
Dynamic Host Configuration Protocol for IPv6 (DHCPv6)   Mike Carney (Sun)
draft-ietf-dhc-dhcpv6-14.txt                            Jim  Bound (Compaq)
Extensions for the DHCP for IPv6                        Charlie Perkins (Nokia)
draft-ietf-dhc-v6exts-11.txt

WG discussed several open issues:

* Releasable resources: IPv6 addresses are one example; are there
  others (none suggested).  When and how should these resources be
  managed (what is leasing strategy) and released (to original server
  or any server)?
* Stateful autoconfiguration and multiple IP addresses: discussion
  centered on tradeoff between complexity and utility - client might
  make multiple requests for addresses in IPv4-IPv6 transition.
  Discussion will continue on DHCPv6 mailing list.
* Stateful autoconfiguration and use of multiple servers (e.g.,
  failover): any useful input from experience with DHCPv4?
* Document organization: authors prefer to leave as two documents.

Authors will pose questions separately to DHCPv6 mailing list for WG
discussion and publish revised draft by end of calendar year.




From dhcp-v4@bucknell.edu  Fri Dec 10 13:35:12 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA00472
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 13:35:11 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA11276;
	Fri, 10 Dec 1999 13:33:59 -0500 (EST)
Date: Fri, 10 Dec 1999 13:33:59 -0500 (EST)
Message-Id: <38513E5C.C91A7E9C@metaip.checkpoint.com>
Errors-To: droms@bucknell.edu
Reply-To: richardj@metaip.checkpoint.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Richard Jones <richardj@metaip.checkpoint.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Configuration Management
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mozilla 4.6 [en] (WinNT; U)
Content-Transfer-Encoding: 7bit

Our product sternly admonishes (;-)) admins to use the UI to do any
kind of configuration or reconfiguration work on the engines. It offers
a remote 'update-restart' mechanism that restarts both members of
a failover relationship, and yelps piteously if the admin restarts one
engine without restarting the other. So, it's a Big Brother approach
that still allows a determined administrator to do whatever they
may choose.

Currently our servers do have to be brought down and brought
back up in order to change their configuration(s).

"Kostur, Andre" wrote:

> I'm curious as to how the rest of the implementors are approaching the
> problem of configuration management between the primary and secondary
> servers.  As far as I can tell, the draft just says (implies?) that the two
> servers are configured identically, and that's that.  However one must still
> deal with the problem.  Does a configuration change mean that both servers
> have to be effectively stopped, apply the configuration change, then restart
> the services?  Or do the two servers talk to each other about config changes
> when they happen?  And in this case, what if there are config changes during
> COMMUNICATIONS-INTERRUPTED or PARTNER-DOWN states (ok, basically any state
> other than NORMAL....)?



From dhcp-v6@bucknell.edu  Fri Dec 10 13:39:12 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA00506
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 13:39:11 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA13737;
	Fri, 10 Dec 1999 13:31:19 -0500 (EST)
Date: Fri, 10 Dec 1999 13:31:19 -0500 (EST)
Message-Id: <21839C593E75D211BF7800805FCCD026EAAEB3@msgsrv03.srv.PacBell.COM>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: "HIBBS, BARR (PB)" <RBHIBBS@msg.pacbell.com>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: RE: Minutes from DC WG meetings
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2448.0)
Mime-Version: 1.0


..ok as is...

--Barr

> ----------
> From: 	Ralph Droms[SMTP:droms@bucknell.edu]
> Sent: 	Friday, December 10, 1999 9:40 AM
> To: 	Multiple recipients of list
> Subject: 	Minutes from DC WG meetings
> 
> Here are the draft minutes from the DC WG meetings.  Please
> review and post comments by this afternoon as the minutes are
> due today.
> 
> I'm sorry for the short notice for review and comment on
> the minutes...
> 
> - Ralph
> 
> Tuesday, Nov 9 at 0900-1130 (DHCPv4)
> ====================================
> 
> Dynamic host configuration : DHCP reconfigure extension Yves T'Joens
> (Alcatel)
> draft-schrijvp-dhcpv4-reconfigure-00.txt
> 
> The author described a proposed option for a unicast message to
> trigger DHCP client reconfiguration.  The WG provided feedback (add
> authentication, make draft more precise) and agreed to consider the
> option after a new draft is submitted.
> 
> 
> The Name Service Search Option for DHCP                 Carl Smith (Sun)
> draft-ietf-dhc-nsso-01.txt
> 
> The WG suggested using a 16-bit value for the service numbers, and to
> define 0 as a reserved value to configure the client to look in local
> configuration table.  The draft will be ready for WG last call after a
> final small revision.
> 
> 
> DHCP Failover Protocol                                  Kim Kinnear
> (Cisco)
> draft-ietf-dhc-failover-05.txt
> 
> Kim reported on phone conferences held since last WG meeting.  He
> will schedule one or two additional phone conferences and submit a new
> draft for (well before?) the next IETF.
> 
> 
> The User Class Option for DHCP                          Jerome Privat (BT)
> draft-ietf-dhc-userclass-04.txt
> 
> The WG provided feedback on the most recent draft: simplify the draft,
> reduce "MUST" policy directives.  The author will revise draft
> according to WG feedback.
> 
> 
> Double Phase DHCP Configuration                         Jerome Privat (BT)
> draft-privat-dhc-doublephase-01.txt
> 
> WG suggested splitting this document into two drafts: an informational
> draft for the proxy mechanism and a standards track draft for redirect
> option.  WG will continue to review this option after new drafts are
> published. 
> 
> 
> Dynamic Registration and Configuration Protocol (DRCP)  Anthony McAuley
> draft-itsumo-drcp-00.txt                                (Telcordia)
> 
> Protocol related to DHCP that reduces DHCP client-server communication
> latency through distributed DHCP servers (or proxies).  The WG
> expressed the opinion that this may be a new protocol, not DHCP.  The
> WG agreed to look at the next draft and decide whether or not to
> review the protocol.
> 
> 
> Authentication for DHCP Messages                        Bill Arbaugh
> (Penn)
> draft-ietf-dhc-authentication-12.txt
> 
> Bill reported on suggestions from Volz. Ran Atkinson suggested
> splitting into two drafts; WG chose to leave "as is".  Draft will go
> to WG last call after next revision.
> 
> 
> DHCP Authentication Via Kerberos V                      Bernard Aboba
> draft-hornstein-dhc-kerbauth-01.txt                     (Microsoft)
> 
> Bernard presented an extension to DHCP to provide Kerberos V
> authentication between clients and servers and to provide interdomain
> authentication.  The basic mechanism is to transport Kerberos
> authentication data in DHCP messages (with, potentially, other
> communication between Kerberos participants).  The WG suggested
> external, authoritative review of the application of Kerberos
> technology and took no specific action.
> 
> 
> DHC load balancing algorithm                            Bernie Volz
> draft-ietf-dhc-loadb-00.txt                             (Process Software)
> 
> This draft results from load balancing technology developed with
> failover protocol, which seemed to have wider application to DHCP
> server configuration.  Mike Patrick expressed concern about the
> 
> necessity of configuring relay agents to match servers (if not, some
> requests may get dropped).  Bernie will revise draft according to
> feedback from WG, and WG will continue to review the document.
> 
> 
> The Server Selection Option for DHCP                    Bill Sommerfield
> draft-ietf-dhc-sso-03.txt                               (Epilogue)
> 
> Rob Stevens suggested this option overlaps with load balancing; WG
> consensus was that overlap was small enough to warrant continued
> consideration of the server selection option.  The WG suggested
> simplifying the draft (e.g., the descriptions of profiles), changing
> several "MUST"s to "SHOULD"s, and to allow lease time as tie breaker.
> Bill will revise and the WG will review the next draft.
> 
> 
> Dynamic Host Configuration Protocol (DHCP) Server MIB   Barr Hibbs
> (Pacbell)
> draft-ietf-dhc-server-mib-04.txt
> 
> Barr will post specific questions to the WG mailing list and revise
> MIB according to feedback.  Next revision should be ready for WG last
> call. 
> 
> 
> Interpreting Client Options for DHCP                    Barr Hibbs
> (Pacbell)
> draft-ietf-dhc-client-options-00.txt
> 
> Barr will withdraw this draft or let it expire.  Ted Lemon volunteered
> to begin work on a revision of RFC 2131/2132 that will be submitted
> for full Standard status.  This revision will include clarifications
> and revised text based on Barr's draft as well as other experience
> with DHCP.
> 
> 
> Option for NAT/IPSec interaction                        Jose Brustoloni
> (no draft available)                                    (Bell Labs)
> 
> This option addresses interoperation problems between NAT and IPSec.
> WG provided feedback and will review a revised draft.
> 
> 
> Subnet selection option                                 Ted Lemon
> draft-ietf-dhc-subnet-option-03.txt                     (for Glenn Waters)
> 
> Draft is ready for WG last call.
> 
> 
> Static routes with subnet masks option                  Ted Lemon
> (no draft available)
> 
> Ted will write up draft for an option to pass a list of static routes
> with subnet masks.
> 
> 
> Domain name search list option                          Ted Lemon
> draft-ietf-dhc-domsrch-02.txt
> 
> 
> Ted will track down previous draft and move it forward.
> 
> 
> Tuesday, Nov 9 at 1300-1400 (DHCPv4)
> ====================================
> 
> DHCP Relay Agent Information Option                     Mike Patrick
> (Motorola)
> draft-ietf-dhc-agent-options-07.txt
> 
> Mike described Motorola IPR statement.  A few new issues were raised:
> subnet sub-option (WG consensus was to keep it); interaction between
> relay agent option and IPSec (relay agent should not add option);
> server should not assume relay agent is present and may choose to add
> a relay agent information option.  The WG agreed to hold a last call
> on a revised document that addresses these points.
> 
> 
> DHCP Schema for LDAP                                    Bernie Volz
> draft-ietf-dhc-schema-01.txt                            (Process Software)
> 
> Changes since last revision:
> 
> * New class, containment and inheritance hierarchy
> * Uses common name attribute
> * Added dhcpPolicy object class and text on mapping dhcpPolicy object
>   to Policy WG schema
> * Removed use of auxiliary classes
> 
> WG agreed to move forward as Proposed Standard document with
> (potentially?) last call after next revision.
> 
> 
> Interaction between DHCP and DNS                        Mark Stapp (Cisco)
> draft-ietf-dhc-dhcp-dns-11.txt
> 
> A new RR for use with DHCP has been proposed.  DHCP will use this
> instead of the previously proposed use of the KEY RR.  Also, the
> security section must be updated to reflect TSIG.  Mark will
> revise and resubmit the draft to reflect these changes.
> 
> 
> A DNS RR for DHCP information                           Andreas Gustafsson
> draft-ietf-dnsind-dhcp-rr-00.txt                        (Internet Engines)
> 
> The DHCP RR (name will likely change) is an opaque data item stored by
> a DNS server and holds a DHCP client identifier to indicate the client
> currently associated with a DNS entry.  The new RR holds only a client
> identifier and no other information (e.g., lease).  The WG consensus
> was that the DHCP RR will meet the requirements for DHCP-DNS
> interaction and should be moved forward.
> 
> 
> New Option Review Guidelines and                        Mike Carney (Sun)
>   Additional Option Namespace
> draft-ietf-dhc-option-review-and-namespace-01.txt
> 
> Mike solicited comments from the WG on section 5 (open issues) and
> would like to go to last call after minor revisions addressing those
> issues.
> 
> 
> Tuesday, Nov 9 at 1415-1515 (DHCPv6)
> ====================================
> Dynamic Host Configuration Protocol for IPv6 (DHCPv6)   Mike Carney (Sun)
> draft-ietf-dhc-dhcpv6-14.txt                            Jim  Bound
> (Compaq)
> Extensions for the DHCP for IPv6                        Charlie Perkins
> (Nokia)
> draft-ietf-dhc-v6exts-11.txt
> 
> WG discussed several open issues:
> 
> * Releasable resources: IPv6 addresses are one example; are there
>   others (none suggested).  When and how should these resources be
>   managed (what is leasing strategy) and released (to original server
>   or any server)?
> * Stateful autoconfiguration and multiple IP addresses: discussion
>   centered on tradeoff between complexity and utility - client might
>   make multiple requests for addresses in IPv4-IPv6 transition.
>   Discussion will continue on DHCPv6 mailing list.
> * Stateful autoconfiguration and use of multiple servers (e.g.,
>   failover): any useful input from experience with DHCPv4?
> * Document organization: authors prefer to leave as two documents.
> 
> Authors will pose questions separately to DHCPv6 mailing list for WG
> discussion and publish revised draft by end of calendar year.
> 
> 


From dhcp-v4@bucknell.edu  Fri Dec 10 15:31:42 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA02680
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 15:31:39 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA30444;
	Fri, 10 Dec 1999 15:29:48 -0500 (EST)
Date: Fri, 10 Dec 1999 15:29:48 -0500 (EST)
Message-Id: <63D30D6E10CFD11190A90000F805FE8601CE700D@lespaul.process.com>
Errors-To: droms@bucknell.edu
Reply-To: Gonczi@process.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Steve Gonczi <Gonczi@process.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2650.21)
Mime-Version: 1.0



> Hold on a sec. With our current setup, let's say the secondary is
> configured via hash bucket assignment to respond to a DISCOVER
> from client A. It assigns an IP, and sometime later client A releases
> it. The address then gets cycled back to the primary. Some time after
> this, client A does either another DISCOVER or a REQUEST, with
> every right to expect the secondary to make all efforts to give it back
> the same IP. However, since the address has been cycled back to
> the primary, and the mechanism for getting addresses back to the
> secondary is based more on whether or not the secondary is
> running low on addresses, I think this design goal goes by the
> wayside. Am I missing something?

Your analysis is correct. One more reason to keep the released/expired 
lease with the backup server, and NOT transferring ownership.

>I guess the only way to deal with this right now is for the secondary
>to issue a POOLREQ every time it relinquishes an IP back to the
>primary. 

We could also trigger an automatic,immediate binding update from 
the primary, setting it back to backup. 

The simplest, obvious solution is to set it back to BACKUP and 
not to FREE.

sG


From dhcp-v4@bucknell.edu  Fri Dec 10 16:42:14 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA03598
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 16:42:14 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA23300;
	Fri, 10 Dec 1999 16:39:41 -0500 (EST)
Date: Fri, 10 Dec 1999 16:39:41 -0500 (EST)
Message-Id: <47FEF5D2BE22D311AA8600508B108915999D21@fs-sd-exch1.coppermountain.com>
Errors-To: droms@bucknell.edu
Reply-To: emichelsen@coppermountain.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Eric Michelsen <emichelsen@coppermountain.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Clearly identifying this mailing list
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2448.0)
Mime-Version: 1.0

Is there a way for the To: field of messages from this exploder to include
the name of the mailing list?  I get mail from lots of sources, and several
exploders, most of which identify themselves in the To: (PPP, DSL Forum,
etc.).  On this list, having the To: field say simply "Multiple recipients
of list" makes it very hard to pick up the thread.
-------------------------------------------
Eric L. Michelsen, Copper Mountain Networks 


From dhcp-v4@bucknell.edu  Fri Dec 10 18:37:26 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id SAA05115
	for <dhc-archive@odin.ietf.org>; Fri, 10 Dec 1999 18:37:18 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id SAA05523;
	Fri, 10 Dec 1999 18:33:17 -0500 (EST)
Date: Fri, 10 Dec 1999 18:33:17 -0500 (EST)
Message-Id: <4250941404.944848779@AD-DELLW2K>
Errors-To: droms@bucknell.edu
Reply-To: ADobkin@emory.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: "Alan S. Dobkin" <ADobkin@emory.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: Clearly identifying this mailing list
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0
X-Mailer: Mulberry (Win32) [2.0.0b4, s/n U-BetaTester]
Content-Transfer-Encoding: 7bit

--On Friday, December 10, 1999 4:43 PM -0500 Eric Michelsen
<emichelsen@coppermountain.com> wrote:

> Is there a way for the To: field of messages from this exploder to include
> the name of the mailing list?  I get mail from lots of sources, and
> several exploders, most of which identify themselves in the To: (PPP, DSL
> Forum, etc.).  On this list, having the To: field say simply "Multiple
> recipients of list" makes it very hard to pick up the thread.
> -------------------------------------------
> Eric L. Michelsen, Copper Mountain Networks 

Actually, the To: header does include the name of the list as part of
the list address, but you are probably using one of those "user friendly" 
e-mail clients that is cutting it off.  Here is the actual header line:

To: Multiple recipients of list <dhcp-v4@bucknell.edu>

Alan

-- 
Alan S. Dobkin, Operating Systems Analyst
Windows NT Group, Enterprise Systems Support
Information Technology Division, Emory University


From dhcp-v4@bucknell.edu  Mon Dec 13 11:43:14 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA11174
	for <dhc-archive@odin.ietf.org>; Mon, 13 Dec 1999 11:43:13 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA09117;
	Mon, 13 Dec 1999 11:42:28 -0500 (EST)
Date: Mon, 13 Dec 1999 11:42:28 -0500 (EST)
Message-Id: <Version.32.19991212165820.00e0ab70@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: droms@bucknell.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Web-based mailing list archive
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

Answers to three questions I've received about the web-based archive:

* Threading does *not* cross month boundaries.

* Searching doesn't work in the prototype but will work in the 
  production version.

* Eventually, access to the raw log files (currently available
  through GET requests to listserv@bucknell.edu) will no longer
  be available.

- Ralph





From dhcp-v6@bucknell.edu  Mon Dec 13 11:47:38 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA11261
	for <dhc-archive@odin.ietf.org>; Mon, 13 Dec 1999 11:47:37 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA13680;
	Mon, 13 Dec 1999 11:37:43 -0500 (EST)
Date: Mon, 13 Dec 1999 11:37:43 -0500 (EST)
Message-Id: <Version.32.19991212165820.00e0ab70@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: RE: Web-based mailing list archive
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

Answers to three questions I've received about the web-based archive:

* Threading does *not* cross month boundaries.

* Searching doesn't work in the prototype but will work in the 
  production version.

* Eventually, access to the raw log files (currently available
  through GET requests to listserv@bucknell.edu) will no longer
  be available.

- Ralph





From dhcp-v4@bucknell.edu  Mon Dec 13 21:06:58 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id VAA22344
	for <dhc-archive@odin.ietf.org>; Mon, 13 Dec 1999 21:06:56 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id VAA07225;
	Mon, 13 Dec 1999 21:05:29 -0500 (EST)
Date: Mon, 13 Dec 1999 21:05:29 -0500 (EST)
Message-Id: <Pine.LNX.4.10.9912140143180.4886-100000@localhost.localdomain>
Errors-To: droms@bucknell.edu
Reply-To: marco@nbnet.nb.ca
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Marco Shaw <marco@nbnet.nb.ca>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: DHCP authentication
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: TEXT/PLAIN; charset=US-ASCII
Mime-Version: 1.0

Is there anyway to authenticate users via a username/password on Windows
clients?

I'm thinking a daemon on the server that checks for a new arp entry with a
MAC address.  The client will still get the IP address along with some
kind of authentication request.  If the client fails, the IP address is
revoked, and the MAC address is 'blacklisted'.  I'm not looking for very
secure method at this time.

Thanks,
Marco




From dhcp-v4@bucknell.edu  Tue Dec 14 02:37:54 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id CAA09461
	for <dhc-archive@odin.ietf.org>; Tue, 14 Dec 1999 02:37:53 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id CAA28399;
	Tue, 14 Dec 1999 02:37:27 -0500 (EST)
Date: Tue, 14 Dec 1999 02:37:27 -0500 (EST)
Message-Id: <47FEF5D2BE22D311AA8600508B108915999D37@fs-sd-exch1.coppermountain.com>
Errors-To: droms@bucknell.edu
Reply-To: emichelsen@coppermountain.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Eric Michelsen <emichelsen@coppermountain.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Clearly identifying this mailing list
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2448.0)
Mime-Version: 1.0

Yes, I see that the emailer I am forced to use is omitting the contents of
the angle brackets, and I see no way to change this broken characteristic.
However, since the label "Multiple recipients of list" adds no information,
how about if it said "DHCP-v4 list" instead?
-------------------------------------------
Eric L. Michelsen, Copper Mountain Networks 

> -----Original Message-----
> From: Alan S. Dobkin [mailto:ADobkin@emory.edu]
> Sent: Friday, December 10, 1999 15:34
> To: Multiple recipients of list
> Subject: Re: Clearly identifying this mailing list
> 
> 
> --On Friday, December 10, 1999 4:43 PM -0500 Eric Michelsen
> <emichelsen@coppermountain.com> wrote:
> 
> > Is there a way for the To: field of messages from this 
> exploder to include
> > the name of the mailing list?  I get mail from lots of sources, and
> > several exploders, most of which identify themselves in the 
> To: (PPP, DSL
> > Forum, etc.).  On this list, having the To: field say 
> simply "Multiple
> > recipients of list" makes it very hard to pick up the thread.
> > -------------------------------------------
> > Eric L. Michelsen, Copper Mountain Networks 
> 
> Actually, the To: header does include the name of the list as part of
> the list address, but you are probably using one of those 
> "user friendly" 
> e-mail clients that is cutting it off.  Here is the actual 
> header line:
> 
> To: Multiple recipients of list <dhcp-v4@bucknell.edu>
> 
> Alan
> 
> -- 
> Alan S. Dobkin, Operating Systems Analyst
> Windows NT Group, Enterprise Systems Support
> Information Technology Division, Emory University
> 


From dhcp-v4@bucknell.edu  Tue Dec 14 11:52:09 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA23841
	for <dhc-archive@odin.ietf.org>; Tue, 14 Dec 1999 11:52:08 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA23898;
	Tue, 14 Dec 1999 11:51:33 -0500 (EST)
Date: Tue, 14 Dec 1999 11:51:33 -0500 (EST)
Message-Id: <199912141636.LAA12462@funnel.cisco.com>
Errors-To: droms@bucknell.edu
Reply-To: droms@bucknell.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: Clearly identifying this mailing list
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

I don't have direct control over the mailing list processor.  I'll ask and
find out if we can arrange to put a better identifier in the "To:" field...

- Ralph

At 02:39 AM 12/14/99 -0500, Eric Michelsen wrote:
>Yes, I see that the emailer I am forced to use is omitting the contents of
>the angle brackets, and I see no way to change this broken characteristic.
>However, since the label "Multiple recipients of list" adds no information,
>how about if it said "DHCP-v4 list" instead?




From dhcp-v4@bucknell.edu  Tue Dec 14 14:10:53 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27034
	for <dhc-archive@odin.ietf.org>; Tue, 14 Dec 1999 14:10:49 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id OAA00100;
	Tue, 14 Dec 1999 14:10:08 -0500 (EST)
Date: Tue, 14 Dec 1999 14:10:08 -0500 (EST)
Message-Id: <199912141834.TAA06750@mephisto.imp.ch>
Errors-To: droms@bucknell.edu
Reply-To: Patrick.Guelat@imp.ch
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Patrick Guelat <Patrick.Guelat@imp.ch>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: dhcp leasequery msgtype
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=US-ASCII
Mime-Version: 1.0
X-Mailer: ELM [version 2.4ME+ PL19 (25)]
Content-Transfer-Encoding: 7bit


I noticed CISCO is using a lot of `new' (draft) dhcp-features in their
cmts-products (Cable modem termination systems, Cisco UBR Series).

They use the agent circuit id option 82 as specified in the draft.
No problem with that one.

But: 

To verify addresses not in the arp table of an interface cisco's 
dhcp server implementation (network registrar 3.01) and the ubr
use a `new' dhcp message type they call 'LEASEQUERY' (type 13).

When the CMTS gets a packet with an unknown source addr on the cable
interface (i.e. that was NOT aquired with dhcp / is not present in
the router's arp-table) the router issues such a 'LEASEQUERY' message
to the dhcp-server.

I traced such a message:

GOT DHCP PACKET WITH LEN 251
        OP    : 1 (BOOTREQUEST)
        HTYPE : 0
        HLEN  : 0
        HOPS  : 0
        XID   : 0
        SECS  : 0
        FLAGS : 0
        CIADDR: 172.16.1.98
        YIADDR: 0.0.0.0
        SIADDR: 0.0.0.0
        GIADDR: 172.16.0.1
        CHADDR: 
        SNAME : 
        FILE  : 

        OPTIONLLEN: 15
        RFC1497 VENDOR MAGIC COOKIE
        OPTION:  53 (DHCP MESSAGE TYPE): 13 - UNKNOWN!!!
        OPTION:  57 (DHCP MAX MSG SIZE): 576
        OPTION:  55 (PARAMETER REQ LIST): 82
        OPTION: 255 (END)

CIADDR seems to be the address to verify.. I have no idea what answer
the UBR expects on this, but it looks like network registrar is
sending DHCPACK and NAK.

Comments ?

	-Patrick


From dhcp-v6@bucknell.edu  Wed Dec 15 17:26:04 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA18612
	for <dhc-archive@odin.ietf.org>; Wed, 15 Dec 1999 17:26:04 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id RAA28269;
	Wed, 15 Dec 1999 17:13:04 -0500 (EST)
Date: Wed, 15 Dec 1999 17:13:04 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.945294971.11328.mwc@atlantic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: general releasable resources? 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0


Thomas writes:

> "HIBBS, BARR (PB)" <RBHIBBS@msg.pacbell.com> writes:
> 
> > However, I think there is one issue we must resolve first:  why must a
> > client request a releasable resource, but not be required to explicitly
> > request non-releasable ones?  If some extensions are intended to always be
> > supplied by a server, are these specified in one of the drafts?  I actually
> > favor a scheme where clients explicitly request all extensions that they
> > desire, with the server supplying any that are required because of [other]
> > protocol requirements or local site policy.  I just want to see the language
> > about what is and which ones are automatically provided and which ones must
> > be requested to be unambiguous in the RFCs.
> 
> My understanding of the reason for requiring that client's explicitly
> request releasable resources is to avoid the problem in DHCPv4 where
> it is problematical for the server to send options that the client
> doesn't understand.

Mostly correct. Clients in DHCPv4 were supposed to ignore options that they
didn't recognize. Unfortunately, some client vendor's implementations fell
over if they got options they didn't request or expect. That left server
vendors grappling with increasingly complicated schemes to control what client
got what options, etc.

My personal opinion is that the original design of the protocol was correct;
clients MUST ignore options they don't recognize/request. Since some
implementors got this wrong, RFC 2131 must not have been clear enough
regarding this point.

v6 is different in that we have been discussing releasable resources in
addition to IP addresses. Since these resources are "releasable", a client
would need to accept some responsibility for managing that resource. Thus, at
Minneapolis, the WG decided that resources which require management by the
client should only be returned if the client requests them, thus communicating
to the server the client's ability to properly "care" for that resource.
Nonreleasable (informational) extensions could either be requested or always
sent depending on the administrator's policy decisions. It's very important
that we get this text right in V6 such that client implementors understand
that they MUST be prepared to receive informational extensions they didn't
request and may not understand, and *ignore* it. > 
> I too share your concern that one of the benefits of DHCP is allowing
> the sys admin to say (via a configuration in the server) here is what
> I want the client to have, independent of what the client actually
> requests. So, maybe what is needed is not that the client explicitely
> request a releasable resource, but a way for the client to indicate
> that it supports a particular resource, should the server choose to
> send it. But even that may be unnecessarily restrictive.

Well, you don't give kittens to kids that aren't responsible enough to care
for them. ;^)

> 
> Thomas



Mike Carney
SNT Internet Engineering
Sun Microsystems, Inc.
Mailstop UBUR02-212
One Network Drive
Burlington, MA 01803




From dhcp-v6@bucknell.edu  Thu Dec 16 11:29:59 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA27893
	for <dhc-archive@odin.ietf.org>; Thu, 16 Dec 1999 11:29:59 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA22687;
	Thu, 16 Dec 1999 11:19:39 -0500 (EST)
Date: Thu, 16 Dec 1999 11:19:39 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.945359149.27374.mwc@atlantic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: DHCPv6 issues status
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0

Hi folks,

The DHCPv6 ID authors (Jim Bound, Charlie Perkins, and myself) meet on
a weekly basis to discuss open issues and the great input we've been
getting from the WG. Thanks much for taking the time to contribute. We
really appreciate it. The following information is based in a large part
on your participation.

Status Update:
==============

1) General Releasable resources

Based upon input from the WG, we've decided to defer the support of
general releasable resources until a later time. We believe we've
uncovered the many if not all of the issues related to general support,
and feel we have a good handle on how the protocol could support such
resources. Later, when the IETF community identifies more releasable
resources, we will revisit this issue. Meanwhile, we would like to
reserve the last 8k options (57344-65535) for assignment of extensions
for dealing with releasable resources.


2) Multiple Addresses, Multiple servers

We feel we've solved the issues of support for multiple addresses (of
different prefixes) and multiple servers. Please read the following
and let us know if you agree.

Acronyms:
=========

ERE	-	Extension Request Extension

General:
========

We started the discussion by talking about scenarios where multiple IP
addresses are consumed, and how the protocol would work in such
environments.

Can we handle many global and/or site prefixes assigned to a link?
==================================================================

Yes. We need a method by which a client can discover what prefixes are
available on the link. This is done as follows:

	- Sophisticated clients MAY include a ERE in their Solicit which
	  requests that servers advertise their prefixes [a].  Servers
	  receiving an ERE with the prefix request will advertise
	  available prefixes on the client's link by including IP
	  address extensions for each prefix, which contains the network
	  address and prefix size (subscriber ID is zero).

	- Since solicits may generate more than one advertisement
	  (multiple servers), a Solicit ID provided in the Solicit (and
	  echoed back in the Advertisement) is used to link
	  advertisements to their solicitations [b].

	- Simple clients do not include the prefix request ERE, and thus
	  only care to get *some* set of one or more IP addresses
	  returned by the server. They make the number of IP addresses
	  known by including empty IP address extensions in their DHCP
	  request.

	- Relay agents forward traffic to remote DHCP servers. DHCP
	  servers have been configured with the mappings from the Relay
	  agent's prefix to available prefixes [c].

	- A client includes the IP address extensions from the
	  advertisement in the Request to signal to the server what
	  flavor IP address(es) it would like.


Requesting additional addresses
===============================

New requests for addresses are differentiated through the use of the XID
cache implemented on the server. A different XID means a new request.
The same means a retransmission. We understand that there is a window
where the XID cache entry may timeout and the server may allocate
additional IP addresses to the client. We manage that problem by
specifying that the XID TIMEOUT be 3 hours by default. We suggest
that somewhat short lifetimes be specified such that IP addresses
allocated which the client does not know about are reclaimed in a
timely fashion. If after this draft goes to PS, and we (WG) find that
this problem proves to be more of a problem than we anticipated, we
will revisit this issue.

3) Router Advertisements

There has been some traffic on the list wondering whether having a
single 'M' bit in the header in router advertisements provided enough
granularity, or whether their should be an 'M' bit in the individual
prefixes. In other words, there may be some problems in environments
where stateless and stateful config are combined. Some folks suggested
that multiple router advertisements would be sent, some for stateless,
others for stateful.

We felt that this issue just wasn't a problem, so long as the stateless
and stateful client implementations operate independently from one
another.

4) Reconfigure/Renumbering

We talked about whether or not we should have a renumbering section in
the appendix. (Yes). We also talked about reconfigure, and whether we
need both schemes or whether we could discard one of them. We've
identified additional semantics of RECONFIGURE, which we'll present in a
separate E-mail message.

5) Authentication

We felt that people calling into question whether or not our
authentication model is adequate given the progress made with DHCPv4
authentication need to raise specific points regarding potential
defects in our specification. Our response is that it is up to those
people to point out the defects to the WG, as we see our model as
having been validated before. "I'm uncomfortable" isn't a helpful
comment - "I'm uncomfortable because of a, b, and c" *is*.

6) API interface for clients

I proposed an informational RFC which would describe an API which could
be used by applications to make requests to and retrieve data from a
DHCPv6 client. The other authors thought this was a good idea. What do
you think?

7) Multiple Interfaces, multiple choices for application layer extensions.

Erik raised a concern about how a client would present a coherent view
of the data returned from multiple interfaces when such data might
conflict. I suggested we use the method the Solaris DHCPv4 client uses
to resolve the conflict - the user picks a "primary" interface, and the
data associated with that interface is authoritative. I've heard no
comment for or against this proposal. Your input is welcomed.


Footnotes, etc
==============

	[a]	ERE needs specification

	[b]	The solicit ID needs specification. Perhaps 8 bits out
		of the reserved field could be used? Is 8 bits enough? A
		solict ID of 0 means that client doesn't match
		advertisements to solicits.

	[c]	This "mapping" seems rather nasty. I'm wondering if we
		should make network administrator's life easier by
		having the relay agent enumerate all of the prefixes
		available on the client's link, rather than force the
		DHCP server administrator to configure this. I think we
		need to do this. Comments?



Mike Carney
SNT Internet Engineering
Sun Microsystems, Inc.
Mailstop UBUR02-212
One Network Drive
Burlington, MA 01803




From dhcp-v6@bucknell.edu  Fri Dec 17 05:15:26 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id FAA01921
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 05:15:26 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id FAA18923;
	Fri, 17 Dec 1999 05:10:15 -0500 (EST)
Date: Fri, 17 Dec 1999 05:10:15 -0500 (EST)
Message-Id: <199912170950.KAA03663@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 issues status 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   1) General Releasable resources
   
=> Address extension should have a type in the reserved area because
it is a releasable resource (ie. its type should be something like 57344).

   2) Multiple Addresses, Multiple servers
   
   	- Simple clients do not include the prefix request ERE, and thus
   	  only care to get *some* set of one or more IP addresses
   	  returned by the server.

=> I don't know if we can be simpler for simple clients: I believe
a simple client will care to get only one address because it is the
simplest thing and the simple client is not smart enough to know
the number of prefixes (what happens when the client askes for too
many addresses?).

          They make the number of IP addresses
   	  known by including empty IP address extensions in their DHCP
   	  request.
   
=> the next I-D should defined what is an empty IP address extension
(zero/undefined address and zero prefix size for instance).

   3) Router Advertisements
   
=> I'd like to get the comment about A bits (ie. A bits are for
stateless autoconf then they should be zero when the M bit is set to one)
in a rationate/explanation statement.

   4) Reconfigure/Renumbering
   
=> reconfigure seems to be the right way to deal with this and
it is an essential new feature of DHCPv6 over DHCP. I am eager to
read the separate E-mail message about it.
   
   6) API interface for clients
   
   I proposed an informational RFC which would describe an API which could
   be used by applications to make requests to and retrieve data from a
   DHCPv6 client. The other authors thought this was a good idea. What do
   you think?
   
=> this is a good idea because DHCPv6 is more complex than DHCP,
for instance we need something like a daemon with a thread per interface
in order to deal with reconfigure (ie. a simple RPC is not enough).

   7) Multiple Interfaces, multiple choices for application layer extensions.
   
   Erik raised a concern about how a client would present a coherent view
   of the data returned from multiple interfaces when such data might
   conflict. I suggested we use the method the Solaris DHCPv4 client uses
   to resolve the conflict - the user picks a "primary" interface, and the
   data associated with that interface is authoritative. I've heard no
   comment for or against this proposal. Your input is welcomed.
   
=> this seems to be a reasonable way to solve this issue (see previous
point too).

Regards

Francis.Dupont@inria.fr

PS: I'd like to have a clarification about how the server recognizes
a client when the client can directly talk to the server without a
relay (ie. how the client fills address fields in Request messages
in such a case and how the server deals with this).
Proposal: the client fills Request messages as it does when it uses
a relay but sends packets directly to the server: client identification
and message delivery are two different things and some fields in
messages are used by both functions and others by only one (as interface
index and prefix-size). This should work but needs clarification.



From dhcp-v4@bucknell.edu  Fri Dec 17 11:16:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA07333
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 11:16:40 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA29059;
	Fri, 17 Dec 1999 11:14:45 -0500 (EST)
Date: Fri, 17 Dec 1999 11:14:45 -0500 (EST)
Message-Id: <5104D4DBC598D211B5FE0000F8FE7EB204087282@mbtlipnt02.btlabs.bt.co.uk>
Errors-To: droms@bucknell.edu
Reply-To: jerome.privat@bt.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: jerome.privat@bt.com
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: draft on DHCP Next Server option
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="ISO-8859-1"
Mime-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2448.0)

As agreed by the WG at the DC meeting, I have split the
DHCP Double Phase Configuration document
(draft-privat-dhc-doublephase-01.txt) into two drafts:
one for the redirect option and one for the proxy approach.

Mike Borella and I produced the draft on the redirect approach.
It can be found at:
http://www.borella.net/draft-ietf-dhc-nextserver-00.txt

It is a generalisation of the option proposed previously which
allows a DHCP server to redirect a client to a secondary configuration
server (DHCP, RSIP, etc...).
As discussed at the meeting, we intend to submit this draft as a
WG document destined for standards-track status (whereas the 
proxy approach will be in an informational document).

Please send your comments to the list.

Thanks,
Jerome Privat,
BT




From dhcp-v6@bucknell.edu  Fri Dec 17 14:01:30 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA10116
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 14:01:29 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA19283;
	Fri, 17 Dec 1999 13:55:18 -0500 (EST)
Date: Fri, 17 Dec 1999 13:55:18 -0500 (EST)
Message-Id: <Roam.SIMC.2.0.6.945454292.20726.mwc@atlantic>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 issues status 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Mime-Version: 1.0


Greetings, Francis -

>  In your previous mail you wrote:
> 
>    1) General Releasable resources
>    
> => Address extension should have a type in the reserved area because
> it is a releasable resource (ie. its type should be something like 57344).

Good point, we might as well be consistent right from the beginning.

> 
>    2) Multiple Addresses, Multiple servers
>    
>    	- Simple clients do not include the prefix request ERE, and thus
>    	  only care to get *some* set of one or more IP addresses
>    	  returned by the server.
> 
> => I don't know if we can be simpler for simple clients: I believe
> a simple client will care to get only one address because it is the
> simplest thing and the simple client is not smart enough to know
> the number of prefixes (what happens when the client askes for too
> many addresses?).

Exactly the point. We want to allow for clients that run the gamut from
very simple (accept what the server gives them) to sophisticated (request
addresses as needed). For the more sophisticated client which asks for too
many addresses, that would be addressed through a policy setting in the server
implementation - max number of addresses per client per network, for example.


> 
>           They make the number of IP addresses
>    	  known by including empty IP address extensions in their DHCP
>    	  request.
>    
> => the next I-D should defined what is an empty IP address extension
> (zero/undefined address and zero prefix size for instance).

Yes, we plan on being explicit in the IP address extension material on how
prefixes are encoded.

> 
>    3) Router Advertisements
>    
> => I'd like to get the comment about A bits (ie. A bits are for
> stateless autoconf then they should be zero when the M bit is set to one)
> in a rationate/explanation statement.

Agreed. We ensure we haven't left any inconsistencies at the interface where
stateless meets stateful. Continuing the discussion thread on this exact issue
would help us accomplish this.

> 
>    4) Reconfigure/Renumbering
>    
> => reconfigure seems to be the right way to deal with this and
> it is an essential new feature of DHCPv6 over DHCP. I am eager to
> read the separate E-mail message about it.

Should go out today, if we (authors) reach consensus on the note.

>    
>    6) API interface for clients
>    
>    I proposed an informational RFC which would describe an API which could
>    be used by applications to make requests to and retrieve data from a
>    DHCPv6 client. The other authors thought this was a good idea. What do
>    you think?
>    
> => this is a good idea because DHCPv6 is more complex than DHCP,
> for instance we need something like a daemon with a thread per interface
> in order to deal with reconfigure (ie. a simple RPC is not enough).

We've found it critical to do this for DHCPv4 as well for our own
applications. It may be worthwhile covering a similar interface for v4 as well
as v6, whether it be in the same document or a separate document.

> 
>    7) Multiple Interfaces, multiple choices for application layer extensions.
>    
>    Erik raised a concern about how a client would present a coherent view
>    of the data returned from multiple interfaces when such data might
>    conflict. I suggested we use the method the Solaris DHCPv4 client uses
>    to resolve the conflict - the user picks a "primary" interface, and the
>    data associated with that interface is authoritative. I've heard no
>    comment for or against this proposal. Your input is welcomed.
>    
> => this seems to be a reasonable way to solve this issue (see previous
> point too).

Thanks for verifying this.

> 
> Regards
> 
> Francis.Dupont@inria.fr
> 
> PS: I'd like to have a clarification about how the server recognizes
> a client when the client can directly talk to the server without a
> relay (ie. how the client fills address fields in Request messages
> in such a case and how the server deals with this).
> Proposal: the client fills Request messages as it does when it uses
> a relay but sends packets directly to the server: client identification
> and message delivery are two different things and some fields in
> messages are used by both functions and others by only one (as interface
> index and prefix-size). This should work but needs clarification.

Yes, the discussion between you and Erik regarding the use of interface indexes
and relay agents wasn't part of this note. We need to address your concerns.

Two things:

1) interface indexes can change if the router is reconfigured differently.
Thus, I don't think they're consistent enough to rely on.

Example: the following is from a machine where I configured le1 first, when
before, le0 was configured first. Note that index 2 is now le1, when before I
rebooted, it was le0.

lo0: flags=1000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv4> mtu 8232 index 1
        inet 127.0.0.1 netmask ff000000 
le1: flags=1004843<UP,BROADCAST,RUNNING,MULTICAST,DHCP,IPv4> mtu 1500 index 2
        inet 192.168.252.251 netmask ffffff00 broadcast 192.168.252.255
        ether 8:0:20:11:df:30 
lo0: flags=2000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv6> mtu 8252 index 1
        inet6 ::1/128 
le0: flags=2000841<UP,RUNNING,MULTICAST,IPv6> mtu 1500 index 3
        ether 8:0:20:11:df:30 
        inet6 fe80::a00:20ff:fe11:df30/10 
le0:1: flags=2080841<UP,RUNNING,MULTICAST,ADDRCONF,IPv6> mtu 1500 index 3
        inet6 fec0:0:ac17:8000:a00:20ff:fe11:df30/64 
le0:2: flags=2080841<UP,RUNNING,MULTICAST,ADDRCONF,IPv6> mtu 1500 index 3
        inet6 2000:0:ac17:8000:a00:20ff:fe11:df30/64
> 

2) Clarification regarding the agent address field never made it into the
current draft. It's the prefix that's sent, not the agent address. The
combination of the prefix, client's link-local address, and one or more IP
address extensions referring to the precise IP address(es) is enough for the
server to identify the client's binding (unless I'm missing something).


Mike Carney
SNT Internet Engineering
Sun Microsystems, Inc.
Mailstop UBUR02-212
One Network Drive
Burlington, MA 01803




From dhcp-v4@bucknell.edu  Fri Dec 17 14:01:54 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA10159
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 14:01:54 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id OAA22945;
	Fri, 17 Dec 1999 14:00:25 -0500 (EST)
Date: Fri, 17 Dec 1999 14:00:25 -0500 (EST)
Message-Id: <199912171815.NAA19187@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: droms@bucknell.edu
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Listserv server unavailable 12/18, 8am-8pm
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

I received the message included below earlier today.  Because
of the mailing list processor upgrades, the dhcp-v4@bucknell.edu,
dhcp-v6@bucknell.edu, netbook@cs.purdue.edu and
netbook-inst@cs.purdue.edu mailing lists will not be available
on 12/18.  I would guess, although I'm not positive, that mail
sent to those mailing lists while the system is down will be
queued and sent after the upgrade is complete.

- Ralph Droms

=====

The Listserv server will be unavailable on Saturday, December 18
from 8am - 8pm while ISR upgrades the listserver for Y2K
compliance and to improve performance.  This means that you will
be unable to use any Bucknell-hosted list to send messages during
this time.  You also might experience infrequent, short
email outages.

After the upgrade you should notice only minor differences,
with one exception.  Any lists whose archives had previously
been available through gopher will now be accessed via a new
web archiving service. For more information,
go to http://www.listproc.bucknell.edu.

If you experience any trouble with a list after the upgrade,
please contact your list owner.  List owners please contact
the Tech Desk @77777 if you experience any problems.

ISR regrets any inconvenience.

Lorraine Eisenhuth
ISR/CCS







--
========================================================================
If you do not wish to receive CCS outage announcements and status
reports via e-mail, you may choose to use any of World Wide Web (WWW),
Usenet Newsgroups or Gopher, instead.  To access the announcements
newsgroup, read the bucknell.ccs.announce newsgroup using the "tin"
program on coral or Newswatcher on a network connected Macintosh.  To
use Gopher, use the "gopher" program on coral or the TurboGopher
program on a network connected Macintosh.  Gopher programs and
newsreaders may also be available on other systems (i.e. the Sun
network).  To use WWW, use your favorite WWW client with the URL:

http://coral.bucknell.edu/cgi-bin/nntp?bucknell.ccs.announce
Informational handouts and training shortcourses are available from CCS.

If you are uninterested in this class of information or wish to use
WWW, gopher or the newsgroup, you may UNSUBSCRIBE from this list by
sending an e-mail message to LISTSERV@BUCKNELL.EDU with a blank subject
and the body of your message containing the line:

UNSUBSCRIBE CCS-ANNOUNCE
========================================================================








From dhcp-v6@bucknell.edu  Fri Dec 17 14:02:52 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA10175
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 14:02:52 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA12256;
	Fri, 17 Dec 1999 13:56:55 -0500 (EST)
Date: Fri, 17 Dec 1999 13:56:55 -0500 (EST)
Message-Id: <199912171815.NAA29577@mail.bucknell.edu>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Ralph Droms <droms@bucknell.edu>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Listserv server unavailable 12/18, 8am-8pm
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.
Content-Type: text/plain; charset="us-ascii"
Mime-Version: 1.0
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0

I received the message included below earlier today.  Because
of the mailing list processor upgrades, the dhcp-v4@bucknell.edu,
dhcp-v6@bucknell.edu, netbook@cs.purdue.edu and
netbook-inst@cs.purdue.edu mailing lists will not be available
on 12/18.  I would guess, although I'm not positive, that mail
sent to those mailing lists while the system is down will be
queued and sent after the upgrade is complete.

- Ralph Droms

=====

The Listserv server will be unavailable on Saturday, December 18
from 8am - 8pm while ISR upgrades the listserver for Y2K
compliance and to improve performance.  This means that you will
be unable to use any Bucknell-hosted list to send messages during
this time.  You also might experience infrequent, short
email outages.

After the upgrade you should notice only minor differences,
with one exception.  Any lists whose archives had previously
been available through gopher will now be accessed via a new
web archiving service. For more information,
go to http://www.listproc.bucknell.edu.

If you experience any trouble with a list after the upgrade,
please contact your list owner.  List owners please contact
the Tech Desk @77777 if you experience any problems.

ISR regrets any inconvenience.

Lorraine Eisenhuth
ISR/CCS







--
========================================================================
If you do not wish to receive CCS outage announcements and status
reports via e-mail, you may choose to use any of World Wide Web (WWW),
Usenet Newsgroups or Gopher, instead.  To access the announcements
newsgroup, read the bucknell.ccs.announce newsgroup using the "tin"
program on coral or Newswatcher on a network connected Macintosh.  To
use Gopher, use the "gopher" program on coral or the TurboGopher
program on a network connected Macintosh.  Gopher programs and
newsreaders may also be available on other systems (i.e. the Sun
network).  To use WWW, use your favorite WWW client with the URL:

http://coral.bucknell.edu/cgi-bin/nntp?bucknell.ccs.announce
Informational handouts and training shortcourses are available from CCS.

If you are uninterested in this class of information or wish to use
WWW, gopher or the newsgroup, you may UNSUBSCRIBE from this list by
sending an e-mail message to LISTSERV@BUCKNELL.EDU with a blank subject
and the body of your message containing the line:

UNSUBSCRIBE CCS-ANNOUNCE
========================================================================








From dhcp-v6@bucknell.edu  Fri Dec 17 17:11:32 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA14006
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 17:11:32 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA19393;
	Fri, 17 Dec 1999 16:47:14 -0500 (EST)
Date: Fri, 17 Dec 1999 16:47:14 -0500 (EST)
Message-Id: <199912172127.WAA00644@givry.inria.fr>
Errors-To: droms@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
Originator: dhcp-v6@bucknell.edu
Sender: dhcp-v6@bucknell.edu
Precedence: bulk
From: Francis Dupont <Francis.Dupont@inria.fr>
To: Multiple recipients of list <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 issues status 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion list for DHCPv6 internet protocol.

 In your previous mail you wrote:

   Two things:
   
   1) interface indexes can change if the router is reconfigured differently.
   Thus, I don't think they're consistent enough to rely on.
   
=> my idea is the relay adds the index of the interface where it has
received the packet from the client then it can use it in order to
forward the answer from the server to the client. The interface index
should not be used for other things (it should be only copied) then
if the relay reboots at most a packet can have a bad interface index.
 I think we need it in the case a prefix is shared between several
interfaces on the relay with only one global address for the whole set
of interfaces (for instance links are "unnumbered" point-to-points,
ie. they have only link-local addresses and one global address
to make DHCPv6 happy. If the relay looks at allocated addresses
it can add a host route per client)
 Of course I no more want to use the interface index for client
identification in the server (cf discussion with Erik).

Thanks

Francis.Dupont@inria.fr


From dhcp-v4@bucknell.edu  Fri Dec 17 19:32:04 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA15916
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 19:32:03 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA22037;
	Fri, 17 Dec 1999 19:30:46 -0500 (EST)
Date: Fri, 17 Dec 1999 19:30:46 -0500 (EST)
Message-Id: <199912180020.QAA03660@scv2.apple.com>
Errors-To: droms@bucknell.edu
Reply-To: cheshire@apple.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Stuart Cheshire <cheshire@apple.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: DHCP Lease Time Option woes
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="iso-8859-1"
Mime-Version: 1.0
X-Mailer: Claris Emailer 2.0v3, January 22, 1998
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by ietf.org id TAA15916

At the DC meeting, Ted Lemon volunteered to collect comments for 
revisions to RFC 2131 as it moves to full standard. Here is one thing I 
think could benefit from some clarification:

The use of the Requested Lease Time Option (51) in DHCP client packets is 
not as obvious as it might seem. There is the question of which packets 
it should appear in, and with what value in each case.

(Remember that I'm coming at this from an Apple perspective -- saying 
that your sysadmin configures it for you is not an acceptable answer for 
us. We want to ship PowerBooks that you can take anywhere in the world, 
and be able to plug them in and have them "just work". If you need a 
sysadmin to come and configure your Lease Time Option then he might as 
well manually configure your IP address too while he's there.)

Here are the five relevant packet types:

DISCOVER
REQUEST (responding to a DHCPOFFER)
REQUEST (when renewing)
REQUEST (when rebinding)
REQUEST (when rebooting)

For each one, I need to work out whether it should include a Lease Time 
Option, and if so, with what value.

Before you assume that it is obvious, and I must be an idiot, bear with 
me. It is more tricky that it seems...

In Open Transport 2.0.x (Mac OS 8.6) OT never put a Lease Time Option in 
any packet. This caused problems with certain DHCP servers, which would 
treat each renewal request as merely being an enquiry of how much time 
remains on the current lease, not a request to actually extend said 
lease. As a result, if you have a one-hour lease, then you renew after 30 
minutes, then after 15 minutes, then after 7.5 minutes, and so on, 
successively halving until the lease expires. The author of the server in 
question quoted RFC 2131 to us:

>   o IF the client has not requested a specific lease in the
>     DHCPDISCOVER message and the client already has an assigned network
>     address, the server returns the lease expiration time previously
>     assigned to that address (note that the client must explicitly
>     request a specific lease to extend the expiration time on a
>     previously assigned address), ELSE

The server author was implementing this very literally. If the client 
already has an assigned network address, and is not explicitly requesting 
a specific lease to extend the expiration time, then the expiration time 
stays at it was.

Suitably chagrined, we modified our DHCP client in OT 2.5.x (Mac OS 9.0) 
to include a specific lease request in all packets. Since we don't 
actually care what the lease duration is, and the whole point of DHCP for 
us is that the parameters are controlled by the server administrator, we 
requested a lease time of "very large" (0x3FFFFFFF) in the assumption 
that the server would cap that large request to the locally configured 
maximum. I deliberately didn't use 0xFFFFFFFF because that is the 
"infinite" lease. I deliberately didn't use 0xFFFFFFFE because I didn't 
want to tickle signed/unsigned bugs in badly written servers. That made 
0x7FFFFFFF the sensible choice, which I halved once more to be extra 
extra cautious. So far, so good.

We went through alpha testing, beta testing, and final candidate testing, 
manufactured lots of Mac OS 9.0 CDs, and started selling them on the 
store shelves.

Then, our Japanese testing office reports that it doesn¹t work with 
Japanese ISDN modems. It turns out that Japanese ISDN modems NAK all 
REQUEST packets with lease request outside their allowed range. This put 
the Mac into an endless DISCOVER/OFFER/REQUEST/NAK deadlock with the 
server.

So, now what are we to do?

If you don¹t request a lease time, RFC 2131 says you don¹t get a lease 
extension.

If you request a very long lease time, then some DHCP servers (claiming 
to follow RFC 1541) NAK you.

If you request a short lease time, then you get a short lease time, 
effectively destroying the ability of a DHCP server administrator to make 
their own decision about what lease duration they wish to hand out to 
clients.

There is no clear answer. I even tried some more subtle workarounds, 
without success. (This is mail is starting to get long, so I'll continue 
with "Stuart's subtle workaround which doesn't work" in a subsequent 
email...)

Stuart Cheshire <cheshire@apple.com>
 * <A HREF="http://ResComp.Stanford.EDU/~cheshire/">Web Page</A>
 * Wizard Without Portfolio, Apple Computer



From dhcp-v4@bucknell.edu  Fri Dec 17 19:33:53 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA15934
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 19:33:53 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA10184;
	Fri, 17 Dec 1999 19:33:41 -0500 (EST)
Date: Fri, 17 Dec 1999 19:33:41 -0500 (EST)
Message-Id: <199912180024.QAA20642@scv3.apple.com>
Errors-To: droms@bucknell.edu
Reply-To: cheshire@apple.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Stuart Cheshire <cheshire@apple.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: DHCP Lease Time Option conclusion
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="US-ASCII"
Mime-Version: 1.0
X-Mailer: Claris Emailer 2.0v3, January 22, 1998

I think that the successor to RFC 2131 needs to be explicit about how the 
server should interpret the presense or absense of lease time options in 
packets from clients:

Either:

(a) Any DHCPDISCOVER or DHCPREQUEST without a lease time option is 
treated as if it contained a request for the server's default lease. In 
other words, *all* DHCPDISCOVER and DHCPREQUEST packets cause the 
creation or sensible renewal of a lease. The server *never* simply offers 
the dregs of a previous lease. If the server is configured to offer 
24-hour leases, then *all* DISCOVERS and REQUESTS that don't explicitly 
request something different get a lease that lasts for 24-hours beginning 
now.

Or, if the ability to do an "informational query" without actually 
extending the lease is considered to be important, then:

(b) All normal DHCPDISCOVER or DHCPREQUEST packets that are not 
"informational queries" MUST include a lease request option. If the 
client has no particular care what the lease time should be, then the 
client needs a way to communicate that. We can specify that a very large 
lease request should be capped to the server's configured maximum, or 
that a request for a lease of zero means a request for the default lease, 
or that a request for a lease of 0xFFFFFFFE (one less than 'infinity') 
means a request for the default lease.

In either case, a client may still request a specific lease time, which 
the server will treat as a hint of what the client would like to see in 
the OFFER or ACK packet. Macs will probably never do this, but I imagine 
that some special purpose devices may know that they only want an address 
for a few seconds.

Any comments?

Stuart Cheshire <cheshire@apple.com>
 * <A HREF="http://ResComp.Stanford.EDU/~cheshire/">Web Page</A>
 * Wizard Without Portfolio, Apple Computer



From dhcp-v4@bucknell.edu  Fri Dec 17 19:35:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id TAA15948
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 19:35:39 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id TAA03580;
	Fri, 17 Dec 1999 19:35:32 -0500 (EST)
Date: Fri, 17 Dec 1999 19:35:32 -0500 (EST)
Message-Id: <199912180023.QAA04119@scv2.apple.com>
Errors-To: droms@bucknell.edu
Reply-To: cheshire@apple.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Stuart Cheshire <cheshire@apple.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: DHCP Lease Time Option continued
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain; charset="US-ASCII"
Mime-Version: 1.0
X-Mailer: Claris Emailer 2.0v3, January 22, 1998

[Continued from "DHCP Lease Time Option woes". You don't need to read 
this unless you are really interested in just how much grief I went 
through trying to get the Mac DHCP client to work with all servers.]

I even tried some more subtle workarounds, without success. What if I 
omit the lease time option in the DISCOVER packet, wait and see what 
lease time the server offers, and then remember that lease time and 
include it in every subsequent DHCPREQUEST?

This makes the Japanese ISDN modems happy, but it doesn't work for other 
reasons:

1. It removes the ability for the server admin to change the lease time. 
Suppose the DHCP server is set to a 30-minute lease. The server admin 
later realizes that this was a mistake, and changes it to 24 hours. 
Unfortunately, all the Macs will have remembered the initial 30-minute 
offer and will continue to request 30-minute leases forever. Even 
rebooting them would not fix the lease lifetime because INIT-REBOOT 
processing would recover the old lease parameters and skip the 
DHCDISCOVER packet that might have allowed them to learn the new lease 
time.

2. Even more subtle, consider this case:

The server admin has configured a 24-hour lease duration.

On Monday, Joe comes to work at 9:00am, plugs in his laptop computer and 
turns it on. It gets a 24-hour lease on an address. By 5:00pm the first 
renew hasn't happened yet. Joe takes his laptop computer home, and plugs 
it in to his home Ethernet. His REQUEST generated in INIT-REBOOT state is 
NAKed because he's on a new subnet now, so his laptop discards the old 
lease information, and gets a new lease from the cable modem, with a 
one-hour duration. At 10:00pm Joe shuts the laptop down and goes to bed.

On Tuesday, Joe comes to work a minute late at 9:01am. His laptop's 
remembered lease information from the cable modem last night has expired, 
so in the morning it begins back in INIT state. It sends a DISCOVER; the 
server has no record of an active lease for this client ID, so it issues 
a new lease, duration 24-hours, expiring tomorrow at 9:01am. So far, so 
good. Joe takes his laptop computer home at 5:00pm, and plugs it in to 
his home Ethernet again, like Monday night.

On Wednesday, Joe gets to work on time. He arrives at 9:00am, and his 
laptop sends a DISCOVER. As explained, the DISCOVER has no lease time 
option. The server DOES have a record of an active lease for this client 
ID, expiring in one minute's time at 9:01am. The RFC 2131 paragraph 
quoted above is very clear about this: "IF the client has not requested a 
specific lease in the DHCPDISCOVER message and the client already has an 
assigned network address, the server returns the lease expiration time 
previously assigned to that address." That means that the remaining time 
on this lease is 60 seconds, which the server puts in its OFFER packet, 
which means that the Mac remembers it and then proceeds to renew its 
lease every 30 seconds for the rest of the day.

So, my workaround of "learn the lease time and remember it" doesn't work 
either.

What are we to do?

Stuart Cheshire <cheshire@apple.com>
 * <A HREF="http://ResComp.Stanford.EDU/~cheshire/">Web Page</A>
 * Wizard Without Portfolio, Apple Computer



From dhcp-v4@bucknell.edu  Fri Dec 17 20:47:24 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA16748
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 20:47:23 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA27855;
	Fri, 17 Dec 1999 20:46:09 -0500 (EST)
Date: Fri, 17 Dec 1999 20:46:09 -0500 (EST)
Message-Id: <21839C593E75D211BF7800805FCCD026EAAEC4@msgsrv03.srv.PacBell.COM>
Errors-To: droms@bucknell.edu
Reply-To: RBHIBBS@msg.pacbell.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: "HIBBS, BARR (PB)" <RBHIBBS@msg.pacbell.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: DHCP Lease Time Option woes
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain;
X-Mailer: Internet Mail Service (5.5.2448.0)
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by ietf.org id UAA16748

> >   o IF the client has not requested a specific lease in the
> >     DHCPDISCOVER message and the client already has an assigned network
> >     address, the server returns the lease expiration time previously
> >     assigned to that address (note that the client must explicitly
> >     request a specific lease to extend the expiration time on a
> >     previously assigned address), ELSE
> 
> 
> If you don¹t request a lease time, RFC 2131 says you don¹t get a lease 
> extension.
> 
..I'm sorry to seem contrary, but your interpretation is just not correct!
If your client already possesses a valid lease, why would you NOT send that
as the requested address in subsequent discover messages?  The solution to
your problem is that simple!

> This caused problems with certain DHCP servers, which would 
> treat each renewal request as merely being an enquiry of how much time 
> remains on the current lease, not a request to actually extend said 
> lease. As a result, if you have a one-hour lease, then you renew after 30 
> minutes, then after 15 minutes, then after 7.5 minutes, and so on, 
> successively halving until the lease expires.
> 
..and some servers continuously extend the lease on subsequent DHCPREQUEST
messages.  Why does the distinction matter to you?  Are you concerned about
processing time-critical data like streaming video?  I can see how a server
which does not extend a lease without receiving a DHCPDISCOVER message might
provoke a brief window when the client cannot receive and process data while
lease re-negotiation is occurring, but DHCPv4 will probably not address that
problem anytime soon.  Maybe you should advise users of time-critical
applications not to use dynamically assigned addresses if the duration of
such streams is likely to exceed any time remaining on a lease.

On the other hand, if you want to press for a fundamental change in the
protocol, that also is okay, just remember how long it may take for that to
occur.

> Then, our Japanese testing office reports that it doesn¹t work with 
> Japanese ISDN modems. It turns out that Japanese ISDN modems NAK all 
> REQUEST packets with lease request outside their allowed range. This put 
> the Mac into an endless DISCOVER/OFFER/REQUEST/NAK deadlock with the
> server.
> 
..the ISDN modem/DHCP server is broken:  as a manufacturer, you're in a
pickle!  This is an unsolvable problem -- for ANY value of lease time that
you pick, I assert that somewhere is a broken server implementation that
thinks it is (1) too long, or (2) too short, or (3) worst of all, if you do
find a value that works with 99.999999999% of all servers, the one for which
it does not work is certain to be the one chosen by your next corporate
purchaser of 100,000 Mac's.  Is a disclaimer acceptable? For example,
"...this version does not operate with the following vendors' DHCP
servers...."

However, I still say that the solution to your problem is simply to specify
the requested IP address when an existing lease is or has been valid....

--Barr





From dhcp-v4@bucknell.edu  Fri Dec 17 20:48:50 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA16770
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 20:48:50 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA06902;
	Fri, 17 Dec 1999 20:48:42 -0500 (EST)
Date: Fri, 17 Dec 1999 20:48:42 -0500 (EST)
Message-Id: <21839C593E75D211BF7800805FCCD026EAAEC5@msgsrv03.srv.PacBell.COM>
Errors-To: droms@bucknell.edu
Reply-To: RBHIBBS@msg.pacbell.com
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: "HIBBS, BARR (PB)" <RBHIBBS@msg.pacbell.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: RE: DHCP Lease Time Option conclusion
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4
Content-Type: text/plain
X-Mailer: Internet Mail Service (5.5.2448.0)
Mime-Version: 1.0


Stuart--

I sympathize with your problem after reading your prior note, but I don't
quite agree with your suggested changes.  There are too many existing server
implementations that pick one or the other of the choices you suggest.  I am
of the opinion that RFC editors cannot possibly know all the circumstances
where a protocol will be used, so trying to legislate one choice over the
other is not likely to succeed.

I'm more inclined to wording such as:

"...DHCPDISCOVER or DHCPREQUEST messages MAY be sent by a client without a
requested lease time option.  If, as specified by local configuration, the
server chooses to reply to these messages it MUST select the
locally-configured default lease time as the lease duration.  Requests for a
zero lease time MUST be construed by the server as a request for the
locally-configured minimum lease time, while requests for a "near-infinite"
lease time (0xFFFFFFFE) MUST be construed by the server as a request for the
locally-configured maximum lease time. Further, even if the client does not
include the lease time option in the parameter request list with a
DHCPREQUEST or DHCPDISCOVER message, the server MUST return that option to
the client in the DHCPACK or DHCPOFFER message."

..and...

"...it is a local administrative policy decision as to whether subsequent
DHCPDISCOVER and/or DHCPREQUEST messages from a client with an existing,
valid lease are to be interpreted as (1) a request for an extension equal to
the original duration of the lease, or (2) a notification that the client is
continuing to use that lease."

I know this is not what you want, but I don't want leases to just be
automatically extended forever!  By changing the wording so that the
decision is removed from the server implementor and give to the server
administrator, I think we both can be satisfied.

You still, however, have the problem with leases possibly expiring during
streaming video feeds.  But at least then you can point to the user and say,
"This is under your control."

--Barr




From dhcp-v4@bucknell.edu  Fri Dec 17 22:26:16 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA18679
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 22:26:16 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id WAA27603;
	Fri, 17 Dec 1999 22:22:27 -0500 (EST)
Date: Fri, 17 Dec 1999 22:22:27 -0500 (EST)
Message-Id: <199912180253.VAA18809@grosse.manhattan.fugue.com>
Errors-To: droms@bucknell.edu
Reply-To: mellon@isc.org
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ted Lemon <mellon@isc.org>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: DHCP Lease Time Option woes 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4


This is a good comment - thanks!

It also explains the message I got on the ISC DHCP mailing list about
OT clients with MacOS 9 not working with some DHCP servers - thanks
for that, too!

To respond directly, in case you're interested:

> If you request a very long lease time, then some DHCP servers (claiming 
> to follow RFC 1541) NAK you.

A DHCP server that follows RFC1541 is not a DHCP server.   The fact
that your client doesn't work with it isn't terribly surprising.
Knowing your story, I think it would be good to phrase RFC2131ng in
such a way that a conforming implementation will not tickle this
interoperability problem, but OTOH what you were trying to do seems
sensible, so I'd say the right thing to do is actually to just respond
to the "bug" report about the ISDN modem with "that's a bug in the
ISDN modem."   Sigh.

> If you request a short lease time, then you get a short lease time, 
> effectively destroying the ability of a DHCP server administrator to make 
> their own decision about what lease duration they wish to hand out to 
> clients.

Here, I would follow what the Microsoft client does: the DHCP server
can return to you a lease longer than the one you requested, and the
Microsoft client will take whatever time comes back in the offer.
The ISC DHCP server can be configured with a min-lease-time parameter
that forces the client to take a longer lease.   This is useful
because the Microsoft client in fact requests a 24-hour lease.

In any case, may I take this opportunity to suggest that you pick up a
copy of the DHCP Handbook, which, while no substitute for the
standard, does actually talk about strategies for dealing with the
very conundrum of which you speak (it just suggests using the
min-lease-time, and completely misses the other points you raised, of
course, since apparently none of us had stumbled over this particular
permutation before).

			       _MelloN_


From dhcp-v4@bucknell.edu  Fri Dec 17 22:27:36 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA18713
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 22:27:36 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id WAA09926;
	Fri, 17 Dec 1999 22:27:30 -0500 (EST)
Date: Fri, 17 Dec 1999 22:27:30 -0500 (EST)
Message-Id: <199912180300.WAA18841@grosse.manhattan.fugue.com>
Errors-To: droms@bucknell.edu
Reply-To: mellon@isc.org
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ted Lemon <mellon@isc.org>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: DHCP Lease Time Option conclusion 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4


> (a) Any DHCPDISCOVER or DHCPREQUEST without a lease time option is 
> treated as if it contained a request for the server's default lease. In 
> other words, *all* DHCPDISCOVER and DHCPREQUEST packets cause the 
> creation or sensible renewal of a lease. The server *never* simply offers 
> the dregs of a previous lease. If the server is configured to offer 
> 24-hour leases, then *all* DISCOVERS and REQUESTS that don't explicitly 
> request something different get a lease that lasts for 24-hours beginning 
> now.

I think this is actually correct - I have no idea what use it serves
to be able to ask how much time the server thinks is left on the
lease.   I'd be curious to know if any clients support this and/or
make use of it.

> (b) All normal DHCPDISCOVER or DHCPREQUEST packets that are not 
> "informational queries" MUST include a lease request option. If the 
> client has no particular care what the lease time should be, then the 
> client needs a way to communicate that. We can specify that a very large 
> lease request should be capped to the server's configured maximum, or 
> that a request for a lease of zero means a request for the default lease, 
> or that a request for a lease of 0xFFFFFFFE (one less than 'infinity') 
> means a request for the default lease.

Actually, I think the Solaris install floppy DHCP client requests an
infinite lease.   I know about this because when it first started
making the rounds, the ISC server started responding with bogus lease
times because I'd done the time arithmetic without accounting for the
possibility of overflow when adding a relative time to an absolute
time.   Needless to say, the ISC server is now more careful about
doing math on lease durations... :')

> Any comments?

Again, thanks for writing your experience up.   I'd like to see this
folded into RFC2131ng somehow!

			       _MelloN_


From dhcp-v4@bucknell.edu  Fri Dec 17 22:38:30 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id WAA18680
	for <dhc-archive@odin.ietf.org>; Fri, 17 Dec 1999 22:26:16 -0500 (EST)
Received: from reef.bucknell.edu (reef.bucknell.edu [134.82.7.6])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id WAA07146;
	Fri, 17 Dec 1999 22:25:20 -0500 (EST)
Date: Fri, 17 Dec 1999 22:25:20 -0500 (EST)
Message-Id: <199912180255.VAA18820@grosse.manhattan.fugue.com>
Errors-To: droms@bucknell.edu
Reply-To: mellon@isc.org
Originator: dhcp-v4@bucknell.edu
Sender: dhcp-v4@bucknell.edu
Precedence: bulk
From: Ted Lemon <mellon@isc.org>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Re: DHCP Lease Time Option continued 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
X-Comment: Discussion of DHCP for IPv4


> I even tried some more subtle workarounds, without success. What if I 
> omit the lease time option in the DISCOVER packet, wait and see what 
> lease time the server offers, and then remember that lease time and 
> include it in every subsequent DHCPREQUEST?

This is actually forbidden - the client is supposed to send the same
options in the DHCPREQUEST as in the DHCPDISCOVER.

			       _MelloN_


From owner-dhcp-v4@bucknell.edu  Sun Dec 19 12:15:57 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA09206
	for <DHC-ARCHIVE@odin.ietf.org>; Sun, 19 Dec 1999 12:15:56 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA01002;
	Sun, 19 Dec 1999 12:14:36 -0500 (EST)
Received: from alcor.process.com (alcor.process.com [192.42.95.16])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id MAA28749
	for <dhcp-v4@bucknell.edu>; Sun, 19 Dec 1999 12:13:51 -0500 (EST)
Received: by process.com (MX V5.1-X A2w8g) id 1; Sun, 19 Dec 1999 12:13:18 -0400
Sender: owner-dhcp-v4@bucknell.edu
Date: Sun, 19 Dec 1999 12:13:17 -0400
From: Bernie Volz <volz@process.com>
To: <dhcp-v4@bucknell.edu>
CC: dhcp-v4@bucknell.edu
Message-ID: <009E2DA2.E58176EB.1@process.com>
Subject: RE: Configuration Management
Reply-To: volz@process.com
X-Sender: volz@process.com
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

The failover draft does imply that both servers have generally the
same configuration. The major items that must agree are the lease pools.
Options sent to clients (such as lease times, etc) could be different
(though clients may thus get different information depending on which
server replies).

For our offering, when an administrator says to apply an updated
configuration via the UI, we trigger both servers to use the new
configuration. It is possible though that one server may not get the
notification (if the network is partitioned, etc).

The protocol needs to be robust enough to handle different
configurations - this may cause a binding update to be rejected because
it is not in one server's pool. Or, the secondary may not be assigned
any addresses for a pool that the primary knows nothing about.

Perhaps we should consider some mechanism for one server to tell the
other that it suspects a misconfiguration. However, how to "recover"
from that (since which server is out of date is not easy to detemrine)
is a good question. I guess one could assume that each server might take
steps to check if the configuration it has is current and use any
updated configuration if not.

Sending over or negotiating configuration information would be a huge
and difficult task as no two server implementations have all of the same
configuration parameters. Limiting this exchange to lease pools might be
reasonable, but again how do you know which server has the correct data?
All you could determine is that the data doesn't match.

The parameters that are critical for making the failover protocol work
(such as agreement on the MCLT) are exchanged (primary's values used).

- Bernie Volz
  Process Software

Date: Thu, 9 Dec 1999 19:03:31 -0500 (EST)
Message-ID: <716D440F8C29D311991100A0C920487423D973@GIDGET>
From: "Kostur, Andre" <Andre@incognito.com>
To: Multiple recipients of list <dhcp-v4@bucknell.edu>
Subject: Configuration Management

I'm curious as to how the rest of the implementors are approaching the
problem of configuration management between the primary and secondary
servers.  As far as I can tell, the draft just says (implies?) that the two
servers are configured identically, and that's that.  However one must still
deal with the problem.  Does a configuration change mean that both servers
have to be effectively stopped, apply the configuration change, then restart
the services?  Or do the two servers talk to each other about config changes
when they happen?  And in this case, what if there are config changes during
COMMUNICATIONS-INTERRUPTED or PARTNER-DOWN states (ok, basically any state
other than NORMAL....)?



From owner-dhcp-v6@bucknell.edu  Mon Dec 20 02:42:03 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id CAA27999
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 20 Dec 1999 02:42:03 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id CAA22903;
	Mon, 20 Dec 1999 02:37:52 -0500 (EST)
Received: from lukla.Sun.COM (lukla.Sun.COM [192.18.98.31])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id CAA16147
	for <dhcp-v6@bucknell.edu>; Mon, 20 Dec 1999 02:37:51 -0500 (EST)
Received: from sunmail1.Sun.COM ([129.145.1.2])
	by lukla.Sun.COM (8.9.3+Sun/8.9.3) with ESMTP id AAA20086
	for <dhcp-v6@bucknell.edu>; Mon, 20 Dec 1999 00:37:50 -0700 (MST)
Received: from jurassic.eng.sun.com (jurassic.Eng.Sun.COM [129.146.83.130])
	by sunmail1.Sun.COM (8.9.1b+Sun/8.9.1/ENSMAIL,v1.6.1-sunmail1) with ESMTP id XAA27396
	for <dhcp-v6@bucknell.edu>; Sun, 19 Dec 1999 23:37:51 -0800 (PST)
Received: from awe174-18 (awe174-18.AWE.Sun.COM [192.29.174.18])
	by jurassic.eng.sun.com (8.9.3+Sun/8.9.3) with SMTP id XAA728600
	for <dhcp-v6@bucknell.edu>; Sun, 19 Dec 1999 23:37:48 -0800 (PST)
Date: Sun, 19 Dec 1999 23:37:30 -0800 (PST)
From: Erik Nordmark <Erik.Nordmark@eng.sun.com>
Reply-To: Erik Nordmark <Erik.Nordmark@eng.sun.com>
Subject: Re: DHCPv6 issues status
To: <dhcp-v6@bucknell.edu>
In-Reply-To: "Your message with ID" <Roam.SIMC.2.0.6.945359149.27374.mwc@atlantic>
Message-ID: <Roam.SIMC.2.0.6.945675450.26966.nordmark@jurassic>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


> 	- Simple clients do not include the prefix request ERE, and thus
> 	  only care to get *some* set of one or more IP addresses
> 	  returned by the server. They make the number of IP addresses
> 	  known by including empty IP address extensions in their DHCP
> 	  request.

Do you have a proposal that can carry these different types of semantics:
 - Client want's no addresses at all (got them from stateless already)
 - Client want's all addresses (using the "default" prefixes) that the
   server hands out
 - Client wants N addresses (even if the server by default hands out M which
   could be more or less than N)

   Erik



From owner-dhcp-v4@bucknell.edu  Mon Dec 20 06:11:39 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id GAA01065
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 20 Dec 1999 06:11:38 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id GAA05145;
	Mon, 20 Dec 1999 06:10:07 -0500 (EST)
Received: from hidiv.cc.itu.edu.tr (hidiv.cc.itu.edu.tr [160.75.2.21])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id GAA20865
	for <dhcp-v4@bucknell.edu>; Mon, 20 Dec 1999 06:09:55 -0500 (EST)
From: akbulut@itu.edu.tr
Received: from aontws4043 (AONTWS4043.cc.itu.edu.tr [160.75.5.43])
	by hidiv.cc.itu.edu.tr (8.9.2/8.9.2) with SMTP id NAA19866
	for <dhcp-v4@bucknell.edu>; Mon, 20 Dec 1999 13:09:12 +0200 (TUR)
Reply-To: <akbulut@itu.edu.tr>
Sender: owner-dhcp-v4@bucknell.edu
To: <dhcp-v4@bucknell.edu>
Subject: help
Date: Mon, 20 Dec 1999 13:08:32 +0200
Message-ID: <46F3C6A28129D311931700105AC5839E0AE0A7@exscc.cc.itu.edu.tr>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-9"
Content-Transfer-Encoding: 8bit
X-Priority: 3 (Normal)
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook 8.5, Build 4.71.2377.0
Importance: Normal
X-MimeOLE: Produced By Microsoft MimeOLE V4.72.3155.0
X-Sender: "Nihal Akbulut" <akbulut@itu.edu.tr>
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN
Content-Transfer-Encoding: 8bit


Hi,
I have a problem..DHCP server seems runnning normal..But some hosts cant get
their reserved ip address..They take a strange address like 169.254.x.x..
I read that it's a new machanism called Automatic Private IP Addressing in
Win98..Ýt says that if a Windows 98 host cant find a DHCP server it
automatically assigns an private ip address (in srange 169.254.x.x) to
itself..And When the DHCP server becomes available again, it will assign
that computer an IP address.
Maybe sometimes hosts cant find the DHCP server,it's okey but I cant
understand that why the computer get it's real address after DHCP became
available..I tried release/renew all but nothing changed..
Can anyone help me?
Thanx..
Nihal Akbulut



From owner-dhcp-v4@bucknell.edu  Mon Dec 20 10:34:01 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA04730
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 20 Dec 1999 10:34:00 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA13058;
	Mon, 20 Dec 1999 10:32:36 -0500 (EST)
Received: from funnel.cisco.com (funnel.cisco.com [161.44.131.24])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id KAA19727
	for <dhcp-v4@bucknell.edu>; Mon, 20 Dec 1999 10:32:21 -0500 (EST)
Received: from kkinnear-nt (ch2-dhcp133-145.cisco.com [161.44.133.145]) by funnel.cisco.com (8.8.5-Cisco.1/8.6.5) with ESMTP id KAA14806; Mon, 20 Dec 1999 10:31:24 -0500 (EST)
Message-Id: <4.2.0.58.19991220101554.01807100@funnel.cisco.com>
X-Sender: kkinnear@funnel.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Mon, 20 Dec 1999 10:32:38 -0500
To: <dhcp-v4@bucknell.edu>
From: Kim Kinnear <kkinnear@cisco.com>
Subject: RE: Configuration Management
Cc: dhcp-v4@bucknell.edu, kkinnear@cisco.com
In-Reply-To: <009E2DA2.E58176EB.1@process.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: kkinnear@cisco.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


Andre, 

I agree completely with Bernie on this.  
I have a couple of comments which I placed in-line, below.

Cheers -- Kim

At 12:13 PM 12/19/99 -0400, Bernie Volz wrote:
>The failover draft does imply that both servers have generally the
>same configuration. The major items that must agree are the lease pools.
>Options sent to clients (such as lease times, etc) could be different
>(though clients may thus get different information depending on which
>server replies).
>
>For our offering, when an administrator says to apply an updated
>configuration via the UI, we trigger both servers to use the new
>configuration. It is possible though that one server may not get the
>notification (if the network is partitioned, etc).
>
>The protocol needs to be robust enough to handle different
>configurations - this may cause a binding update to be rejected because
>it is not in one server's pool. Or, the secondary may not be assigned
>any addresses for a pool that the primary knows nothing about.

         Yes, this is the key point.  The protocol has to keep
         operating in the face of unsynchronized configuration
         changes between partners.  Although you can always stop
         them from communicating, make the config changes, and
         then restart the communications.  We have not had to do
         that with our implementation, and have had no problems
         traceable to any configuration changes.


>Perhaps we should consider some mechanism for one server to tell the
>other that it suspects a misconfiguration. 

         To some (very small) degree we do have this: one of the
         NAK reply codes is "Illegal IP address", which is
         designed to highlight an IP address not part of any pool
         on the receiving server.  If a message is received with
         an IP address that a server doesn't know about,
         presumably it is a configuration error -- since the
         sending server shouldn't have sent it if it didn't think
         the receiving server was going to be able to respond to
         it.

>However, how to "recover"
>from that (since which server is out of date is not easy to detemrine)
>is a good question. I guess one could assume that each server might take
>steps to check if the configuration it has is current and use any
>updated configuration if not.
>
>Sending over or negotiating configuration information would be a huge
>and difficult task as no two server implementations have all of the same
>configuration parameters. Limiting this exchange to lease pools might be
>reasonable, but again how do you know which server has the correct data?
>All you could determine is that the data doesn't match.

         I tried this in an early "interserver" draft, and it was
         one (among, unfortunately, several) reason why that draft
         more or less fell over under its own weight.  There is a
         very strong Working Group belief that including
         configuration information in the failover protocol is a
         bad idea, no doubt partly as a result of this early attempt
         at doing so.

>The parameters that are critical for making the failover protocol work
>(such as agreement on the MCLT) are exchanged (primary's values used).
>
>- Bernie Volz
>   Process Software
>
>Date: Thu, 9 Dec 1999 19:03:31 -0500 (EST)
>Message-ID: <716D440F8C29D311991100A0C920487423D973@GIDGET>
>From: "Kostur, Andre" <Andre@incognito.com>
>To: Multiple recipients of list <dhcp-v4@bucknell.edu>
>Subject: Configuration Management
>
>I'm curious as to how the rest of the implementors are approaching the
>problem of configuration management between the primary and secondary
>servers.  As far as I can tell, the draft just says (implies?) that the two
>servers are configured identically, and that's that.  However one must still
>deal with the problem.  Does a configuration change mean that both servers
>have to be effectively stopped, apply the configuration change, then restart
>the services?  Or do the two servers talk to each other about config changes
>when they happen?  And in this case, what if there are config changes during
>COMMUNICATIONS-INTERRUPTED or PARTNER-DOWN states (ok, basically any state
>other than NORMAL....)?
>



From owner-dhcp-v4@bucknell.edu  Mon Dec 20 13:46:02 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA11618
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 20 Dec 1999 13:46:02 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA12882;
	Mon, 20 Dec 1999 13:44:15 -0500 (EST)
Received: from toccata.fugue.com (toccata.fugue.com [204.152.188.66])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id NAA04126
	for <dhcp-v4@bucknell.edu>; Mon, 20 Dec 1999 13:43:41 -0500 (EST)
Received: from grosse.manhattan.fugue.com (tundruk.manhattan.fugue.com [160.79.19.71]) by toccata.fugue.com (8.9.3/8.6.11) with ESMTP id KAA28731; Mon, 20 Dec 1999 10:38:33 -0800 (PST)
Received: from grosse.manhattan.fugue.com (localhost [127.0.0.1]) by grosse.manhattan.fugue.com (8.9.3/8.6.11) with ESMTP id NAA00762; Mon, 20 Dec 1999 13:43:36 -0500 (EST)
Message-Id: <199912201843.NAA00762@grosse.manhattan.fugue.com>
To: <dhcp-v4@bucknell.edu>
cc: dhcp-v4@bucknell.edu
Subject: Re: help 
In-Reply-To: Message from akbulut@itu.edu.tr 
   of "Mon, 20 Dec 1999 13:08:32 +0200." <46F3C6A28129D311931700105AC5839E0AE0A7@exscc.cc.itu.edu.tr> 
Date: Mon, 20 Dec 1999 13:43:36 -0500
From: Ted Lemon <mellon@isc.org>
Reply-To: mellon@isc.org
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


This means that for some reason, your DHCP client is unable to contact
the DHCP server.   You need to use a network sniffer to figure out why
that is - maybe the network card on the machine isn't working, maybe
the cable isn't working, maybe a switch or router isn't working, or
maybe the DHCP server itself isn't working.

BTW, this mailing list is not intended as a place to get support for
DHCP servers.   If you are using the ISC DHCP server, please subscribe
to the ISC DHCP mailing list (see the README file in the DHCP
distribution for details).   If you are not using the ISC DHCP server,
please contact your vendor for support or subscribe to the
dhcp-interest mailing list, which is hosted by the ISC but is not
ISC-specific.   You can subscribe by sending mail to
dhcp-interest-request@isc.org.   Then send your actual request for
further assistance to dhcp-interest@isc.org (you can't do this without
subscribing to the mailing list).

			       _MelloN_



From owner-dhcp-v4@bucknell.edu  Mon Dec 20 15:44:25 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA14621
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 20 Dec 1999 15:44:25 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA15032;
	Mon, 20 Dec 1999 15:43:06 -0500 (EST)
Received: from mail.ptd.net (mail1.ha-net.ptd.net [207.44.96.65])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA19293
	for <dhcp-v4@bucknell.edu>; Mon, 20 Dec 1999 15:42:54 -0500 (EST)
Received: (qmail 28752 invoked from network); 20 Dec 1999 20:42:26 -0000
Received: from host22.bvtc.ptd.net (HELO droms-laptop) (209.173.2.22)
  by mail.ptd.net with SMTP; 20 Dec 1999 20:42:26 -0000
Message-Id: <Version.32.19991220150034.00e1f4d0@mail.bucknell.edu>
X-Sender: droms@mail.bucknell.edu
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.2 
Date: Mon, 20 Dec 1999 20:00:58 -0500
To: <dhcp-v4@bucknell.edu>
From: Ralph Droms <droms@bucknell.edu>
Subject: I-D ACTION:draft-ietf-dhc-migagntlist-00.txt
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed
Reply-To: droms@bucknell.edu
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

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

	Title		: The Migration Agent List Option for DHCP
	Author(s)	: N. Kulshreshtha
	Filename	: draft-ietf-dhc-migagntlist-00.txt
	Pages		: 4
	Date		: 10-Dec-99
	
This document defines a new DHCP option for delivering configuration
information to hosts utilizing the Compatibility Mode Driver (CMD)
provided with Novell's NetWare 5. This option is passed from the DHCP
server to the DHCP Client to provide the addresses of one or more
Migration Agents along with their CMD network number.

A URL for this Internet-Draft is:




http://www.ietf.org/internet-drafts/draft-ietf-dhc-migagntlist-00.txt

Internet-Drafts are also available by anonymous FTP. Login with the
username "anonymous" and a password of your e-mail address. After
logging in, type "cd internet-drafts" and then

	"get draft-ietf-dhc-migagntlist-00.txt".

A list of Internet-Drafts directories can be found in


http://www.ietf.org/shadow.html or


ftp://ftp.ietf.org/ietf/1shadow-sites.txt


Internet-Drafts can also be obtained by e-mail.

Send a message to:
	mailserv@ietf.org.
In the body type:
	"FILE /internet-drafts/draft-ietf-dhc-migagntlist-00.txt".
	
NOTE:	The mail server at ietf.org can return the document in
	MIME-encoded form by using the "mpack" utility.  To use this
	feature, insert the command "ENCODING mime" before the "FILE"
	command.  To decode the response(s), you will need "munpack" or
	a MIME-compliant mail reader.  Different MIME-compliant mail readers
	exhibit different behavior, especially when dealing with
	"multipart" MIME messages (i.e. documents which have been split
	up into multiple messages), so check your local documentation on
	how to manipulate these messages.
		
		
Below is the data which will enable a MIME compliant mail reader
implementation to automatically retrieve the ASCII version of the
Internet-Draft.
Content-Type: text/plain
Content-ID:	<19991210114959.I-D@ietf.org>

ENCODING mime
FILE /internet-drafts/draft-ietf-dhc-migagntlist-00.txt



From owner-dhcp-v6@bucknell.edu  Mon Dec 20 17:53:10 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA18221
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 20 Dec 1999 17:53:10 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id RAA06686;
	Mon, 20 Dec 1999 17:49:07 -0500 (EST)
Received: from mailhost.iprg.nokia.com (mailhost.iprg.nokia.com [205.226.5.12])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id RAA32625
	for <dhcp-v6@bucknell.edu>; Mon, 20 Dec 1999 17:49:05 -0500 (EST)
Received: from darkstar.iprg.nokia.com (darkstar.iprg.nokia.com [205.226.5.69]) by mailhost.iprg.nokia.com (8.8.8/8.6.10) with ESMTP id OAA20343; Mon, 20 Dec 1999 14:48:33 -0800 (PST)
Received: (from root@localhost)
	by darkstar.iprg.nokia.com (8.9.3/8.9.3-VIRSCAN) id OAA05791;
	Mon, 20 Dec 1999 14:48:32 -0800
X-Virus-Scanned:  Mon, 20 Dec 1999 14:48:32 -0800 Nokia Silicon Valley AntiVirus Appliance
Received: from <charliep@iprg.nokia.com> (charliep.iprg.nokia.com [205.226.2.89]) by darkstar.iprg.nokia.com  SMTP/WTS (12.69)
 xma005573; Mon, 20 Dec 99 14:48:27 -0800
Sender: owner-dhcp-v6@bucknell.edu
Message-ID: <385EB23B.EA874632@iprg.nokia.com>
Date: Mon, 20 Dec 1999 14:48:27 -0800
From: "Charles E. Perkins" <charliep@IPRG.NOKIA.COM>
Organization: Nokia Research Center
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
X-Accept-Language: en
MIME-Version: 1.0
To: <dhcp-v6@bucknell.edu>
CC: dhcp-v6@bucknell.edu
Subject: [Fwd: [Fwd: Re: DHCPv6 issues status]]
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: dhcp-v6@bucknell.edu
X-Sender: charliep@iprg.nokia.com
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN
Content-Transfer-Encoding: 7bit



Hello Erik,

Thanks for your note.

Here are my simple answers.  There are actually more cases, but I
don't want to get into every detail just yet.  I think Mike is
preparing a more comprehensive note with additional details on
various subjects.

>> 	- Simple clients do not include the prefix request ERE, and thus
>> 	  only care to get *some* set of one or more IP addresses
>> 	  returned by the server. They make the number of IP addresses
>> 	  known by including empty IP address extensions in their DHCP
>> 	  request.
>
> Do you have a proposal that can carry these different types of semantics:
>  - Client want's no addresses at all (got them from stateless already)

Client doesn't include any IP address extension or ERE with address
extension number.

>  - Client want's all addresses (using the "default" prefixes) that the
>    server hands out

Client includes an ERE or IP address extension, probably with no IP
address
(also implies no prefix) in it.

>  - Client wants N addresses (even if the server by default hands out M which
>    could be more or less than N)

Client includes an IP address extension.
If it gets M (<N) addresses, it asks again.
If it gets K (>N) addresses, it releases some.

The last case is pretty speculative, and I don't think it should
be a common case.  Thus I don't see that we need to devise special
protocol to handle it.


Regards,
Charlie P.



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 10:54:41 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id KAA26253
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 10:54:40 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id KAA00420;
	Tue, 21 Dec 1999 10:51:35 -0500 (EST)
Received: from funnel.cisco.com (funnel.cisco.com [161.44.131.24])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id KAA28843
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 10:51:20 -0500 (EST)
Received: from kkinnear-nt (ch2-dhcp133-145.cisco.com [161.44.133.145]) by funnel.cisco.com (8.8.5-Cisco.1/8.6.5) with ESMTP id KAA11608; Tue, 21 Dec 1999 10:50:35 -0500 (EST)
Message-Id: <4.2.0.58.19991221102712.0165a3b0@funnel.cisco.com>
X-Sender: kkinnear@funnel.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Tue, 21 Dec 1999 10:51:53 -0500
To: <dhcp-v4@bucknell.edu>
From: Kim Kinnear <kkinnear@cisco.com>
Subject: Re: DHCP Lease Time Option conclusion 
Cc: Multiple recipients of list <dhcp-v4@bucknell.edu>, kkinnear@cisco.com
In-Reply-To: <199912180300.WAA18841@grosse.manhattan.fugue.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: kkinnear@cisco.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


Stuart,

You bring up some interesting questions.

My comments are inline, related to Ted's reply, below.

At 10:27 PM 12/17/99 -0500, Ted Lemon wrote:

> > (a) Any DHCPDISCOVER or DHCPREQUEST without a lease time option is 
> > treated as if it contained a request for the server's default lease. In 
> > other words, *all* DHCPDISCOVER and DHCPREQUEST packets cause the 
> > creation or sensible renewal of a lease. The server *never* simply offers 
> > the dregs of a previous lease. If the server is configured to offer 
> > 24-hour leases, then *all* DISCOVERS and REQUESTS that don't explicitly 
> > request something different get a lease that lasts for 24-hours beginning 
> > now.
>
>I think this is actually correct - I have no idea what use it serves
>to be able to ask how much time the server thinks is left on the
>lease.   I'd be curious to know if any clients support this and/or
>make use of it.

I agree with Ted.  I too think this is correct.  I am somewhat
embarrassed to note that the paragraph where you don't get an
extension in a DISCOVER unless you asked for the specific IP
address was a surprise to me.  Seems a very strange overloading.

That said, I would object to revised language in the RFC that
said a server MUST offer an entirely new lease period to any
DHCPDISCOVER for several reasons -- I think the language should
be SHOULD, not MUST.

(The reasons are: 1) If two DHCPDISCOVERs come pretty close
together and the lease has almost all of its term still to go
before it expires, we'll offer the remaining lease time as
something of an optimization, and 2) failover causes all lease
time offers to be subject to failover correctness constraints.)


> > (b) All normal DHCPDISCOVER or DHCPREQUEST packets that are not 
> > "informational queries" MUST include a lease request option. If the 
> > client has no particular care what the lease time should be, then the 
> > client needs a way to communicate that. We can specify that a very large 
> > lease request should be capped to the server's configured maximum, or 
> > that a request for a lease of zero means a request for the default lease, 
> > or that a request for a lease of 0xFFFFFFFE (one less than 'infinity') 
> > means a request for the default lease.

I think this is the wrong approach.  If the client doesn't care,
I think it should not include a lease time option.  I do like the
specification of how to ask for the max and min values, though.

Cheers -- Kim


  



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 11:35:41 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA16736
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 11:35:40 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA09841;
	Tue, 21 Dec 1999 11:34:26 -0500 (EST)
Received: from funnel.cisco.com (funnel.cisco.com [161.44.131.24])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA23468
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 11:34:12 -0500 (EST)
Received: from kkinnear-nt (ch2-dhcp133-145.cisco.com [161.44.133.145]) by funnel.cisco.com (8.8.5-Cisco.1/8.6.5) with ESMTP id LAA14430; Tue, 21 Dec 1999 11:33:38 -0500 (EST)
Message-Id: <4.2.0.58.19991221105636.0198b260@funnel.cisco.com>
X-Sender: kkinnear@funnel.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Tue, 21 Dec 1999 11:07:34 -0500
To: <dhcp-v4@bucknell.edu>
From: Kim Kinnear <kkinnear@cisco.com>
Subject: RE: Failover protocol, POOLREQUEST issues
Cc: dhcp-v4@bucknell.edu, kkinnear@cisco.com
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: kkinnear@cisco.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


Steve, et.  al.,

I've been thinking about the POOLREQUEST issues that you have
brought up.

First, my previous analysis was flawed, as you pointed out, in
that it would take two BNDUPD/BNDACK sequences for an expired
address to transit from the secondary to the primary and back
again today, and only one with some of the proposals that have
been kicked around recently.

I've spent some time trying to figure out what problem we are
trying to solve here.  There seems to be an implicit assumption
that an expiring (or released) lease on an IP address that the
secondary server leased out would "naturally" want to flow back
to the secondary -- and thus the concept of ownership of leased
IP addresses that you have proposed.

I see two cases here:

   1.  In the general case, where a lease frees up on the
   secondary that was originally leased on the secondary, I don't
   think there is a problem to solve -- at least then.  This is
   because after the lease was originally granted (and long before
   it frees up), the primary and secondary servers are going to
   have to have replenished the secondary's store of available
   leases.  The shortage of available IP addresses on the
   secondary is an event widely separated in time from the actual
   lease expiration (generally).  There is no telling a priori
   which server needs a newly available IP address more -- the
   primary or secondary, and it certainly would have no
   relationship to who originally leased it.

   2.  In the load balancing case, there would be some value in
   having an expired or released IP address end up as part of the
   available pool of the server who most recently leased it to a
   client, since there is good reason to try to give that client
   back the same IP address if it shows up asking for an IP
   address again.  The loadbalancing algorithm will ensure that it
   shows up at the same server the second time, so it makes sense
   for the same server to have it available.

Of course, if that server fails, then that client isn't going to
get the same IP address back if it asks while the other server,
the only one running, is in COMMUNICATIONS-INTERRUPTED state, but
that is ok -- at least it gets *some* IP address.

---------------------

So I do see two problems to solve (perhaps not exactly the ones
originally giving rise to your proposal):

    1.  Keeping the available pools of each server balanced in the
    way that they have been configured.  This needs activity at
    two discrete times:

	a.  Whenever either server leases out an IP address, some
	rebalancing of the available pools might need to occur.

	b.  Whenever either server senses a lease expiration or
	release, there is the question of what to do with it.

    2.  At time (b) above, trying to ensure that the client that
    used it last can get it again.

I'll make a proposal to solve these problems in a separate mail.

Cheers -- Kim



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 11:39:20 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA23189
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 11:39:20 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA24078;
	Tue, 21 Dec 1999 11:39:07 -0500 (EST)
Received: from funnel.cisco.com (funnel.cisco.com [161.44.131.24])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA26266
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 11:38:54 -0500 (EST)
Received: from kkinnear-nt (ch2-dhcp133-145.cisco.com [161.44.133.145]) by funnel.cisco.com (8.8.5-Cisco.1/8.6.5) with ESMTP id LAA14702; Tue, 21 Dec 1999 11:38:20 -0500 (EST)
Message-Id: <4.2.0.58.19991221110042.01986450@funnel.cisco.com>
X-Sender: kkinnear@funnel.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Tue, 21 Dec 1999 11:39:39 -0500
To: <dhcp-v4@bucknell.edu>
From: Kim Kinnear <kkinnear@cisco.com>
Subject: POOLREQ proposal for failover
Cc: kkinnear@cisco.com, dhcp-v4@bucknell.edu
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: kkinnear@cisco.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


Steve,

These are what I see as the problems to solve concerning
balancing available pools among failover servers: (from my
previous mail on the subject):

    1.  Keeping the available pools of each server balanced in the
    way that they have been configured.  This needs activity at
    two discrete times:

	a.  Whenever either server leases out an IP address, some
	rebalancing of the available pools might need to occur.

	b.  Whenever either server senses a lease expiration or
	release, there is the question of what to do with it.

    2.  At time (b) above, trying to ensure that the client that
    used it last can get it again.

----------------------

I propose the following...

[I wanted to propose junking the POOLREQ, but realized that the
one useful purpose served by the POOLREQ is to signal one server
that its partner has updated it with all relevant information.
This is an issue because it is easy to have one server try to
make available pool balancing decisions prior to it having enough
information to make cogent decisions on the subject.  The POOLREQ
is one way for one server to inform its partner that the partner
now *has* sufficient information to make such a decision.]

Assume the existing protocol mechanisms with slight extensions:
the POOLREQ can be used the secondary to get the primary to
balance its available pools of IP addresses.  And, the BNDUPD can
be used by either server to attempt to "get" or to "give" any
available IP address from or to the other server.

These mechanisms would be used as follows:

1.  Whenever the secondary enters NORMAL state (including the
     first time):

   The secondary would send a POOLREQ to the primary after it has
   updated the primary with all outstanding information (if any)
   that it has the the primary didn't.  [Just like in the current
   draft.]

   The primary would scan all of its available IP address pools,
   and figure out how many IP addresses it should send to the
   secondary.  It would then look at the loadbalancing
   configuration (if any), and send the secondary available
   addresses that had been used most recently by clients that the
   secondary is configured to support.  If the addresses found
   using that technique do not satisfy the needs of the secondary
   server, the primary would send the least recently used IP
   addresses to the secondary.

   The algorithm used here is the same as that used in #3 below,
   except that it is run over *all* of the address pools in which
   these servers are jointly participating in failover.  [This is
   similar to what is in the draft today, with the addition of
   knowledge of the loadbalancing algorithm and sending to the
   secondary IP addresses that match its client base under
   loadbalancing.]

2.  Whenever a lease was granted in NORMAL state:

   The server granting a lease would check the number of available
   addresses it had in the pool from which the lease was granted,
   and how many its partner had.  When the number it needs to meet
   the configuration guidelines increases above a certain
   threshold, it will ask the other server for some specific
   available IP addresses using BNDUPD.  It will ask first for any
   which were recently used by clients which it supports (via
   loadbalancing), if any, and then for the lease recently used IP
   addresses.  [This is all new, not in the current draft.]

3.  Whenever a lease expires or is released in NORMAL state:

   The server processing the expiration or release will compare
   the number of available IP addresses it has in this pool with
   the number of IP addresses that its partner has in this pool,
   and if the difference is above some threshold, it will send
   some available IP addresses to its partner.

   It will send first any IP addresses which were recently used by
   clients which its partner supports (via loadbalancing), if any,
   and then those least recently used.

   It may or may not send the currently expiring or released IP
   address to the partner -- it depends on which server was
   configured to process the most recent client's MAC address due
   to loadbalancing.  [This is all new, not in the current draft.]

In any of the above cases, the fact that a client has recently
used an IP address and that one server or the other is configured
due to loadbalancing to support that client will be considered
when making allocation decisions between servers.  However, the
first priority is to balance the pools of available IP addresses
(within some threshold), and this balance should take priority
over which server is configured to process the client that most
recently leased the IP address.

------------------------

I *believe* that this proposal solves the problems that I've
stated (which flow from, but are not identical to, the problems
that Steve brought up).

Comments?

Thanks -- Kim



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 14:39:19 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA27006
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 14:39:18 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id OAA15119;
	Tue, 21 Dec 1999 14:37:34 -0500 (EST)
Received: from lespaul.process.com (lespaul.process.com [192.42.95.27])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id OAA26575
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 14:37:14 -0500 (EST)
Received: by lespaul.process.com with Internet Mail Service (5.5.2650.21)
	id <ZFB9RA21>; Tue, 21 Dec 1999 14:36:44 -0500
Message-ID: <63D30D6E10CFD11190A90000F805FE8601CE7026@lespaul.process.com>
From: Steve Gonczi <Gonczi@process.com>
To: <dhcp-v4@bucknell.edu>
Subject: re: Failover protocol, POOLREQUEST issues
Date: Tue, 21 Dec 1999 14:36:42 -0500
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Reply-To: Gonczi@process.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

Kim says:

>There seems to be an implicit assumption
>that an expiring (or released) lease on an IP address that the
>secondary server leased out would "naturally" want to flow back
>to the secondary 

Yes. As the protocol is written today, they do. 


>   1.  In the general case,...

Do you mean the case _without_ load balancing? 
I have no issue with that at all. Works fine as-is.

>   2.  In the load balancing case, there would be some value in
>   having an expired or released IP address end up as part of the
>   available pool of the server who most recently leased it to a
>   client, since there is good reason to try to give that client
>   back the same IP address ...

Absolutely. 


>    1.  Keeping the available pools of each server balanced in the
>    way that they have been configured.  This needs activity at
>    two discrete times:
>
>	a.  Whenever either server leases out an IP address, some
>	rebalancing of the available pools might need to occur.

I am not convinced this is worth doing. It reminds me of the centrally
planned
economies of the Soviet block. You would have to be pretty smart about
rebalancing , including looking ahead to see what expirations are expected
in the near future to have a hope of making things better instead of worse. 
(Mind you, there is no way to foresee how many clients are going to boot 
or release in the next minute).

Imagine this: Oops, I am low on leases. Give me some. Oops, forget that, 
a whole bunch just freed up. 

Overall, you may be doing a lot of work, to end up in a situation that is 
slightly worse than it would be without your attempts to fix it.

It may make sense to make pool percentage adjustments based on some
statistics, over longer periods of time. (Intelligent self-tuning).

>	b.  Whenever either server senses a lease expiration or
>	release, there is the question of what to do with it.

Seems obvious to me. Keep it with the server who had it in the first place.
Why not?



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 15:04:17 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA27572
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 15:04:16 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA03391;
	Tue, 21 Dec 1999 15:04:04 -0500 (EST)
Received: from lespaul.process.com (lespaul.process.com [192.42.95.27])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id PAA12438
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 15:04:03 -0500 (EST)
Received: by lespaul.process.com with Internet Mail Service (5.5.2650.21)
	id <ZFB9RAK1>; Tue, 21 Dec 1999 15:03:32 -0500
Message-ID: <63D30D6E10CFD11190A90000F805FE8601CE7027@lespaul.process.com>
From: Steve Gonczi <Gonczi@process.com>
To: <dhcp-v4@bucknell.edu>
Cc: "'dhcp-v4@bucknell.edu'" <dhcp-v4@bucknell.edu>
Subject: RE: Failover protocol, POOLREQUEST issues
Date: Tue, 21 Dec 1999 15:03:26 -0500
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Reply-To: Gonczi@process.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

>>Yes. As the protocol is written today, they do. (go back to the primary)
>
>         No, they flow back to the primary, then on next POOLREQ they
>         go wherever they are needed.  Or did I miss something
>         in what you said?


 I am talking about what happens during normal operation. ( Both servers
are up). Are you assuming periodic POOLREQ -s in normal mode?
My assumption: we are still debating if that is the right solution.


>         My point is that doing what you suggest does not even begin
>         to handle the issue of actually balancing the pools on the two
>         servers, so that if one goes down, then the other may not have
>         enough IP addresses to handle the outage.

Well, there is no way to foresee which one goes down. Hence the best bet 
is to split the lease pools evenly to begin with.

Sugsequently, the "randomness" of client transactions and expirations
are probably as good a way of maintaining the 50-50 as any balancing
algoritm we could come up with. 

>         Leases run for hours to days to weeks, and when they expire
>         there is no reason to think that the machine which originated
>         the lease is necessarily the one that needs more IP addresses.
         
>         Cheers -- Kim

Like I said, one server may need addresses this second, but that may change
the next second. Hence, any re-balancing you attempt, is likely 
going to make things worse. 

sG



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 15:06:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA27607
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 15:06:38 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA01544;
	Tue, 21 Dec 1999 15:06:23 -0500 (EST)
Received: from lespaul.process.com (lespaul.process.com [192.42.95.27])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id PAA03246
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 15:06:13 -0500 (EST)
Received: by lespaul.process.com with Internet Mail Service (5.5.2650.21)
	id <ZFB9RAKP>; Tue, 21 Dec 1999 15:05:42 -0500
Message-ID: <63D30D6E10CFD11190A90000F805FE8601CE7028@lespaul.process.com>
From: Steve Gonczi <Gonczi@process.com>
To: <dhcp-v4@bucknell.edu>
Cc: "'dhcp-v4@bucknell.edu'" <dhcp-v4@bucknell.edu>
Subject: RE: POOLREQ proposal for failover
Date: Tue, 21 Dec 1999 15:05:39 -0500
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Reply-To: Gonczi@process.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

Kim, 

This seems a bit complicated. 
If you want to define a self-tuning heuristic, we should make it 
optional and define it in a separate paragraph in the draft.
Seems like a lot of work for a marginal benefit.

Consider, that load balancing is to some degree self-tuning.
If both servers are running, and one server is out of addresses in one
or more pools, and the other server has such addresses, then
new clients that need these will be served by the other server.

sG



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 15:32:15 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA00052
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 15:32:14 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA06083;
	Tue, 21 Dec 1999 15:30:58 -0500 (EST)
Received: from funnel.cisco.com (funnel.cisco.com [161.44.131.24])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id PAA16427
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 15:30:53 -0500 (EST)
Received: from kkinnear-nt (ch2-dhcp133-145.cisco.com [161.44.133.145]) by funnel.cisco.com (8.8.5-Cisco.1/8.6.5) with ESMTP id PAA29302; Tue, 21 Dec 1999 15:30:20 -0500 (EST)
Message-Id: <4.2.0.58.19991221151733.0197f740@funnel.cisco.com>
X-Sender: kkinnear@funnel.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Tue, 21 Dec 1999 15:31:39 -0500
To: <dhcp-v4@bucknell.edu>
From: Kim Kinnear <kkinnear@cisco.com>
Subject: RE: POOLREQ proposal for failover
Cc: "'Kim Kinnear'" <kkinnear@cisco.com>,
        "'dhcp-v4@bucknell.edu'" <dhcp-v4@bucknell.edu>
In-Reply-To: <63D30D6E10CFD11190A90000F805FE8601CE7028@lespaul.process.c
 om>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: kkinnear@cisco.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

At 03:05 PM 12/21/99 -0500, Steve Gonczi wrote:
>Kim, 
>
>This seems a bit complicated. 
>If you want to define a self-tuning heuristic, we should make it 
>optional and define it in a separate paragraph in the draft.
>Seems like a lot of work for a marginal benefit.

         I view this proposal as considerably less complicated (in
         practice, not necessarily in description) than the one
         that you made several weeks ago, where servers "own" IP
         addresses even when they are leased.

         If the benefit is truly marginal (something of which I'm
         not convinced), then let's leave the draft the way it is.
         I thought you brought up your proposal because you saw a
         problem with pools getting out of balance.  Perhaps I've
         misunderstood your motivation.


>Consider, that load balancing is to some degree self-tuning.
>If both servers are running, and one server is out of addresses in one
>or more pools, and the other server has such addresses, then
>new clients that need these will be served by the other server.

         I don't see this at all.  My understanding of the
         loadbalancing algorithm and its relationship to failover
         is that it defines exactly which MAC addresses each
         server is allowed to respond to for DHCPDISCOVER requests
         when the servers are in NORMAL state.

         It isn't in any way self-tuning, at least as I understand
         it.  If a server is out of IP addresses, then it doesn't
         respond, and the other server *can't* respond as long as
         it is in NORMAL state.  Is there something I missed about
         the algorithm that makes this untrue?

         I quote from draft-ietf-dhc-loadb-00.txt:

         Section 4, paragraph 5:

           "Resource allocation is accomplished by assigning
           specific hash values to each participating server.

           Each server is assigned ownership of one or more hash
           values, and will only service requests where the STID
           hash matches one of these values.

           Any hash buckes not assigned to servers will result in
           some client ST-s being entirely ignored.  (In some
           scenarios, this may be the desired outcome.)"

         Now, I think it would be *nice* it were self-tuning
         with respect to pool size.  I just don't think it
         is now.

         Thanks -- Kim




>sG



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 15:34:34 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA03538
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 15:34:34 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA09232;
	Tue, 21 Dec 1999 15:34:17 -0500 (EST)
Received: from funnel.cisco.com (funnel.cisco.com [161.44.131.24])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id PAA03877
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 15:34:02 -0500 (EST)
Received: from kkinnear-nt (ch2-dhcp133-145.cisco.com [161.44.133.145]) by funnel.cisco.com (8.8.5-Cisco.1/8.6.5) with ESMTP id PAA29493; Tue, 21 Dec 1999 15:33:29 -0500 (EST)
Message-Id: <4.2.0.58.19991221153228.0165c360@funnel.cisco.com>
X-Sender: kkinnear@funnel.cisco.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.2.0.58 
Date: Tue, 21 Dec 1999 15:34:47 -0500
To: <dhcp-v4@bucknell.edu>
From: Kim Kinnear <kkinnear@cisco.com>
Subject: re: Failover protocol, POOLREQUEST issues
Cc: <dhcp-v4@bucknell.edu>, kkinnear@cisco.com
In-Reply-To: <63D30D6E10CFD11190A90000F805FE8601CE7026@lespaul.process.c
 om>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Reply-To: kkinnear@cisco.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

At 02:36 PM 12/21/99 -0500, Steve Gonczi wrote:
>Kim says:
>
> >There seems to be an implicit assumption
> >that an expiring (or released) lease on an IP address that the
> >secondary server leased out would "naturally" want to flow back
> >to the secondary 
>
>Yes. As the protocol is written today, they do. 

          No, they flow back to the primary, then on next POOLREQ they
          go wherever they are needed.  Or did I miss something
          in what you said?


> >   1.  In the general case,...
>
>Do you mean the case _without_ load balancing? 
>I have no issue with that at all. Works fine as-is.
>
> >   2.  In the load balancing case, there would be some value in
> >   having an expired or released IP address end up as part of the
> >   available pool of the server who most recently leased it to a
> >   client, since there is good reason to try to give that client
> >   back the same IP address ...
>
>Absolutely. 
>
>
> >    1.  Keeping the available pools of each server balanced in the
> >    way that they have been configured.  This needs activity at
> >    two discrete times:
> >
> >       a.  Whenever either server leases out an IP address, some
> >       rebalancing of the available pools might need to occur.
>
>I am not convinced this is worth doing. It reminds me of the centrally
>planned
>economies of the Soviet block. You would have to be pretty smart about
>rebalancing , including looking ahead to see what expirations are expected
>in the near future to have a hope of making things better instead of worse. 
>(Mind you, there is no way to foresee how many clients are going to boot 
>or release in the next minute).
>Imagine this: Oops, I am low on leases. Give me some. Oops, forget that, 
>a whole bunch just freed up. 
>
>Overall, you may be doing a lot of work, to end up in a situation that is 
>slightly worse than it would be without your attempts to fix it.
>
>It may make sense to make pool percentage adjustments based on some
>statistics, over longer periods of time. (Intelligent self-tuning).

          That is my point of having a threshold in the proposal that
          I sent.

> >       b.  Whenever either server senses a lease expiration or
> >       release, there is the question of what to do with it.
>
>Seems obvious to me. Keep it with the server who had it in the first place.
>Why not?
>

          My point is that doing what you suggest does not even begin
          to handle the issue of actually balancing the pools on the two
          servers, so that if one goes down, then the other may not have
          enough IP addresses to handle the outage.

          Leases run for hours to days to weeks, and when they expire
          there is no reason to think that the machine which originated
          the lease is necessarily the one that needs more IP addresses.
          
          Cheers -- Kim



From owner-dhcp-v4@bucknell.edu  Tue Dec 21 20:05:51 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id UAA16326
	for <DHC-ARCHIVE@odin.ietf.org>; Tue, 21 Dec 1999 20:05:49 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id UAA02459;
	Tue, 21 Dec 1999 20:04:33 -0500 (EST)
Received: from lespaul.process.com (lespaul.process.com [192.42.95.27])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id UAA17008
	for <dhcp-v4@bucknell.edu>; Tue, 21 Dec 1999 20:04:19 -0500 (EST)
Received: by lespaul.process.com with Internet Mail Service (5.5.2650.21)
	id <ZFB9RA6C>; Tue, 21 Dec 1999 20:03:48 -0500
Message-ID: <63D30D6E10CFD11190A90000F805FE8601CE7029@lespaul.process.com>
From: Steve Gonczi <Gonczi@process.com>
To: <dhcp-v4@bucknell.edu>
Cc: "'dhcp-v4@bucknell.edu'" <dhcp-v4@bucknell.edu>
Subject: RE: POOLREQ proposal for failover
Date: Tue, 21 Dec 1999 20:03:45 -0500
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-Type: text/plain;
	charset="iso-8859-1"
Reply-To: Gonczi@process.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

>         I view this proposal as considerably less complicated (in
>         practice, not necessarily in description) than the one
>         that you made several weeks ago, where servers "own" IP
>         addresses even when they are leased.

Let me give it a second look. I got a bit carried away.

>         I thought you brought up your proposal because you saw a
>         problem with pools getting out of balance.  Perhaps I've
>         misunderstood your motivation.

I was concerned about the backup server simply running out of leases,
unless we introduced some way of replenishment while in normal mode
(periodic POOLREQ-s, retaining ownership or something along the 
lines of your proposal.)

>         I don't see this at all.  My understanding of the
>         loadbalancing algorithm and its relationship to failover
>         is that it defines exactly which MAC addresses each
>         server is allowed to respond to for DHCPDISCOVER requests
>         when the servers are in NORMAL state.

You are absolutely right. For a moment, I lost sight of what happens when 
a specific client is not able to get a lease.  

At one point, we have discussed making the loadbalancing calculation into a
bias 
instead of a hard yes/no. e.g.: we could allow a server to respond to a
DISCOVER even if the hash says it should not, based on the 'secs' field, or 
after the client sent the DISCOVER several times. 
That could make loadbalancing self-tuning ...?
My current thinking is that load balancing should only be applied to
DISCOVER-s.

sG



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 11:12:39 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA27516
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 11:12:39 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA24128;
	Thu, 23 Dec 1999 11:03:05 -0500 (EST)
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA21758
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:03:04 -0500 (EST)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA21556
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:03:02 -0500
Received: from rotala.raleigh.ibm.com (rotala.raleigh.ibm.com [9.37.82.31])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA03452
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:03:02 -0500
Received: from rotala.raleigh.ibm.com (localhost [127.0.0.1]) by rotala.raleigh.ibm.com (8.9.3/8.7/RTP-ral-1.0) with ESMTP id LAA08881 for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:01:35 -0500
Message-Id: <199912231601.LAA08881@rotala.raleigh.ibm.com>
To: <dhcp-v6@bucknell.edu>
Subject: Re: [Fwd: [Fwd: Re: DHCPv6 issues status]] 
In-Reply-To: Message from "Charles E. Perkins" <charliep@iprg.nokia.com> 
   of "Mon, 20 Dec 1999 14:48:27 PST." <385EB23B.EA874632@iprg.nokia.com> 
Date: Thu, 23 Dec 1999 11:01:35 -0500
From: Thomas Narten <narten@raleigh.ibm.com>
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

"Charles E. Perkins" <charliep@iprg.nokia.com> writes:

> >  - Client wants N addresses (even if the server by default hands out M which
> >    could be more or less than N)

> Client includes an IP address extension.
> If it gets M (<N) addresses, it asks again.
> If it gets K (>N) addresses, it releases some.

> The last case is pretty speculative, and I don't think it should
> be a common case.  Thus I don't see that we need to devise special
> protocol to handle it.

This last case is the most interesting one, and is the one that has
caused the confusion/complexity so far. This is one of the critical
aspects of the "request more addresses" model.  Saying it won't be a
common case and we don't need to worry about it now seems odd. The
spec needs to nail down the details and semantics of this type of
operation precisely because they are non-trivial.

Alternatively, if it really is the case that this won't be a common
scenario, why should the protocol support it?

Thomas



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 11:17:13 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA27636
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 11:17:13 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA32295;
	Thu, 23 Dec 1999 11:16:24 -0500 (EST)
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA04151
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:16:15 -0500 (EST)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA06284
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:16:14 -0500
Received: from rotala.raleigh.ibm.com (rotala.raleigh.ibm.com [9.37.82.31])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA22050
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:16:15 -0500
Received: from rotala.raleigh.ibm.com (localhost [127.0.0.1]) by rotala.raleigh.ibm.com (8.9.3/8.7/RTP-ral-1.0) with ESMTP id LAA08919 for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:14:48 -0500
Message-Id: <199912231614.LAA08919@rotala.raleigh.ibm.com>
To: <dhcp-v6@bucknell.edu>
Subject: DHCPv6 address allocation model
In-Reply-To: Message from Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> 
   of "Thu, 16 Dec 1999 11:19:58 EST." <Roam.SIMC.2.0.6.945359149.27374.mwc@atlantic> 
Date: Thu, 23 Dec 1999 11:14:48 -0500
From: Thomas Narten <narten@raleigh.ibm.com>
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> writes:

> We feel we've solved the issues of support for multiple addresses (of
> different prefixes) and multiple servers. Please read the following
> and let us know if you agree.

Re: multiple addresses. I posted a note to the dhcpv6 list back in
august (appended below). Does your solution solve all the points
mentioned in it? Indeed, has the solution been explained to the WG
yet?

Thomas

From: Thomas Narten <narten@hygro.adsl.duke.edu>
To: dhcp-v6@bucknell.edu
Date: Sat, 21 Aug 1999 08:20:10 -0400
Subject: DHCPv6 address allocation model

One of the questions I've had in the back of my mind for a while is
how does one use DHCPv6 to renumber a site. This may seem like a silly
and obvious question, but I'm actually quite serious. To summarize, I
believe there are a lot of ambiguities and unspecified details
relating to the new DHCPv6 semantics, where a client is allowed to ask
for additional addresses. In the following, I describe a number of
concrete scenarios where I am not sure how the protocol is supposed to
work and what the spec calls for. 

Background:

Using stateless address autoconfiguration, the system administrator
controls the prefixes that routers advertise via RAs. Hosts
autoconfigure addresses from the prefixes in RAs. A host generates
exactly one address for each prefix. To renumber a site, the sys admin
adds a new prefix to the list sent out in RAs (which all the hosts
then generate an address for) and also deprecates the (now old) prefix
that has been advertised in use for some time, but now needs to be
phased out. Because hosts always generate exactly one address for each
advertised prefix, there is never any ambiguity as to which addresses
are actually in use, or how they are used.

DHCPv4 doesn't have the notion of muliple addresses per client.
However, it does have the model that the client asks for its
resources, and the server returns them. If the client asks again, it
gets the same answer again. The v6 counterpart of DHCPv4 that would
retain this model (but incorporate v6's multiple address model) would
simply have the server return ALL the addresses a client is supposed
to have (i.e., addresses from the same set of prefixes the RAs would
be advertising). If a client asked for addresses three different
times, it would always get the same answer.

With DHCPv6 (as currently defined), however, the address assignment
model is quite different. Specifically, a server doesn't simply
respond saying "here are ALL of your addresses." If a client asks for
addresses (and the server returns some), and then later asks for more,
the server will return more (i.e., different ones in addition to those
it has already handed out). The motivation for this model (from what I
understand) is to allow individual applications to have their own
addresses.  This new model leads to a number of details that I don't
think are covered in the current spec. The following gives concrete
examples where there are gaps in the current spec, and where I'm not
sure how things are actually supposed to work.

1) In a request, the client must specifically ask for an address (via
an address extension). When it requests an address, how many does the
server return? The text doesn't seem to say, and in fact says:

> 3.2.2. Receiving a DHCP Request with the IP Address Extension
> 
>    When a server receives a request for an IP address, it consults its
>    allocation tables and determines an IP address appropriate for the
>    requesting client and the link to which the client is attached.

note "determines an IP address", i.e., just one.

2) Now, the above wording could of course be tweaked to say "one or
more ...". Let's assume that the above wording is changed and that a
site has 2 prefixes (for two ISPs), and the DHCP server is allowed to
return two addresses in response to the request for an address. No
problem.

But what if the client actually requests two addresses (by including
two address extensions in the request)? How many addresses does the
server return? Which ones?

It could return 2, but that may be incorrect. If the two requests
(i.e., two address extensions included in the request) were for two
different uses, shouldn't the server return 4? Two addresses from each
ISP prefix?

If the server returns 4, what will the client do with those four
addresses? It asked for 2, but got 4. Does it assume the first two
should be associated with the first requested address, while the
latter two belong to the second request? Or does it assume the first
address goes with the first request, the remaining addresses go with
the last request? What if the server returns the 4 addresses in the
different order than the client expects them to be grouped (i.e., the
first two addresses for ISP 1, the last two for ISP 2)?  The point
here is that there are quite a lot of unspecified details in how this
will work. If the client and server have different assumptions, the
wrong thing might happen and we won't have interoperability.

3) DHCPv4 and IPv6's stateless address autoconfiguration have an
assumption that the sys admin knows how many addresses a host
needs. The model is the client can ask for its addresses, and the
server (DHCPv4 server or router sending RAs) always responds giving it
all of the addresses its supposed to use.

The current DHCPv6 model has the client requesting one or more
addresses. But this begs a fundamental question. How does a client
know how many addresses it should ask for? A client generally doesn't
know how many ISPs it has connections through (and there is no defined
mechanism in IPv6 today for a client to find out) so how can it
sensibly be expected to know how many to ask for?

Indeed, how can it ask indicate *which* type of address it wants?  For
example, how would it indicate that it wants two addresses and that
the two addresses to come from two different ISP prefixes?

Another way of stating the issue is that DHCP is intended for
configuring clients. That is, the client is assumed to need
configuring. But if the client is expected to ask the server for
individual addresses, it needs to know what types of addresses to ask
for (e.g., global scope or site-local?, which ISP prefixes, etc.) But
this type of information is normally managed by the centralized DHCP
server. The client doesn't (and shouldn't) need to know these sorts of
things in order to use DHCP. DHCP's primary benefit comes when clients
are dumb, so to speak. For example, when a new machine plugs into the
network the first time, it should be able to use DHCP and just have
things work, without requiring advance configuration of the client.

4) Suppose a site has two prefixes for ISPA and ISPB. A client has
requested and received addresses for both prefixes (i.e., ADDR-A and
ADDR-B). Suppose now that the client sends in a Request to renew
address ADDR-A (the address for ISPA), but doesn't include ADDR-B in
the request. What does the server return? Just an address for ADDR-A
with a new lifetime?  Should it also return an address corresponding
to ISPB? In case 1) above, the answer was yes. But in this case,
returning an additional address from ISPB's prefix gets
complicated. Should it return the same ADDR-B it returned earlier
(that the client is using, but didn't include in an extension in the
Request message). Should the server return ADDR-BPRIME, an address for
ISPB but different from ADDR-B? I'm not sure there is a "right" answer
here. The right answer depends on how the client intends to use the
returned addresses.

5) Suppose a site has one prefix for ISPA, but decides it wants to add
a new one for a new ISP connection (ISPB). If a client had asked for
one address, this wouldn't be too big of a problem. The server could
send a reconfigure, the client would send a Request, and the Reply
could contain both the old address and the new one.

But suppose further that a client has requested and received 2
separate addresses already (both addresses are on prefix ISPA). The
server can send a reconfigure for both addresses, and the client then
responds with a Request for those two addresses. But now we have the
problem described earlier about how many prefixes get returned, which
ones, and what does a client do with them.

6) Suppose a client asks for 4 addresses (two each from ISP prefixes
ISPA and ISPB) but gets back only 2 because the prefix from ISPA is no
longer valid. What are semantics?

In summary, I think there are a number of technical problems with the
current definition of address allocation. I believe that they all stem
from the extended allocation model that is new to DHCPv6. No doubt,
they can all be fixed in the protocol, but I continue to wonder
whether the effort is worth while. The WG should consider:

1) what are the benefits of the new model, and do the benefits justify
   the complexity they add.

   My understanding is that the main benefit is that it allows
   individual applications to request addresses for their own
   use. I.e., so I can have (say) multiple web server processes, each
   using their own address that is presumably not shared with any
   other rapplications.

   My sense also is that there is no concrete example that calls for
   this problem to be solved via DHCP.

   It also appears that the new allocation model has introduced
   complexity to the protocol (relative to DHCPv4) that stems directly
   from the new model. Specifically:

   - 'C' bit is needed so that the a client has a way to tell the
     server to throw away previous allocations. This is needed because
     the client can forget (i.e., via system restart) what resources
     have been assigned to it, and the server has no way of
     determining whether a client is still using a resource. The 'C'
     bit opens up a denial of service attack not present in DHCPv4.
     
   - the requirement that the  server  cache responses and respond to
     retransmissions from its cache for certain time periods. Unlike
     DHCPv4, DHCPv6 requests are not idempotent.

   - Additional complexity will likely be needed to address  some
     of the issues outlined above.
   

2) What additional pieces are also needed in order take advantage of
   the new facilities. I believe that there is additional stuff that
   needs to be done. For example, an API would probably be needed that
   allowed applications to interact with DHCP to request addresses,
   and also that would allow DHCP to interact with an application. For
   example, if an application wants to get an address for its own use,
   it needs to ask the local DHCP entity to obtain one. Likewise, if
   the DHCP server sends out a RECONFIGURE, and that reconfigure
   corresponds to a particular address being used by one particular
   application, there would need to be system hooks (e.g., via the
   API) that allowed the client DHCP process to inform the application
   that the address it was using has changed, a new one has been
   assigned, etc. Because RECONFIGUREs can arrive at any time, the API
   would probably need some sort of asynchronous notification
   facility. (I.e, an application would need to poll or DHCP would
   need do upcalls.)

   Is there agreement that such an API required in order to take
   advantage of these facilities? Does anyone have any plans to write
   one or think that they might in the future?

3) In the case where a client needs to do specialized things (like
   host multiple servers), another argument is that that DHCP provides
   only limited benefits. I.e., such environments will require that
   individual clients be hand-configured anyway, in which case
   configuring the address-specific stuff might make sense anyway (and
   wouldn't be a significant burden), and the incremental benefit of
   using DHCP might be quite small.

Thomas



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 11:33:12 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA28078
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 11:33:11 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA07476;
	Thu, 23 Dec 1999 11:32:01 -0500 (EST)
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA15012
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:31:50 -0500 (EST)
Received: from rtpmail02.raleigh.ibm.com (rtpmail02.raleigh.ibm.com [9.37.172.48])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA33120
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:31:49 -0500
Received: from rotala.raleigh.ibm.com (rotala.raleigh.ibm.com [9.37.82.31])
	by rtpmail02.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA24258
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:31:49 -0500
Received: from rotala.raleigh.ibm.com (localhost [127.0.0.1]) by rotala.raleigh.ibm.com (8.9.3/8.7/RTP-ral-1.0) with ESMTP id LAA08965 for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:30:22 -0500
Message-Id: <199912231630.LAA08965@rotala.raleigh.ibm.com>
To: <dhcp-v6@bucknell.edu>
Subject: DHCPv6 and multiple servers
In-Reply-To: Message from Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> 
   of "Thu, 16 Dec 1999 11:19:58 EST." <Roam.SIMC.2.0.6.945359149.27374.mwc@atlantic> 
Date: Thu, 23 Dec 1999 11:30:22 -0500
From: Thomas Narten <narten@raleigh.ibm.com>
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> writes:

> We feel we've solved the issues of support for multiple addresses (of
> different prefixes) and multiple servers. Please read the following
> and let us know if you agree.

I'd like to be sure we understand what is meant by "multiple servers".
The description supplied in your note seems to cover the case where
there are multiple servers, and the client wants to pick the one of
them for interacting with, i.e., to get all of its parameters from
that one. The current draft already handles this, and the rest of your
note is consistent with that I believe.

But there were also some other thoughts expressed on the about what
"multiple servers" means. There are (at least) two cases. In the first
case, the client has two interfaces and thus runs DHCP on both of
them. Your proposed solution addresses this case. In this case, it is
assumed that the client must run two instances of DHCP with (possibly)
different servers. But the assumption here is that the client gets
information from both sources, and accepts the union of the
information. The fact that it is interacting with two DHCP servers is
an artifact of the the fact that it has two interfaces. No changes to
the DHCP protocol/model are needed to handle this.

There was another case discussed, where a client has one interface,
but may want to interact with multiple servers *simultaneously*.  By
interact, I mean actually get some parameters from one server, other
parameters from another. Is your position that this second case should
not be pursued?  (I personally think that not handling this case is
reasonable, but haven't quite gotten a sense of what the WG thinks
here.) If this case is to be pursued, however, I believe explicit text
needs to go into the document explaining when this is desirable, how a 
client decides it needs to interact with multiple servers, etc.

Thomas

Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> writes:

> Hi folks,

> The DHCPv6 ID authors (Jim Bound, Charlie Perkins, and myself) meet on
> a weekly basis to discuss open issues and the great input we've been
> getting from the WG. Thanks much for taking the time to contribute. We
> really appreciate it. The following information is based in a large part
> on your participation.

> Status Update:
> ==============

> 1) General Releasable resources

> Based upon input from the WG, we've decided to defer the support of
> general releasable resources until a later time. We believe we've
> uncovered the many if not all of the issues related to general support,
> and feel we have a good handle on how the protocol could support such
> resources. Later, when the IETF community identifies more releasable
> resources, we will revisit this issue. Meanwhile, we would like to
> reserve the last 8k options (57344-65535) for assignment of extensions
> for dealing with releasable resources.


> 2) Multiple Addresses, Multiple servers

> We feel we've solved the issues of support for multiple addresses (of
> different prefixes) and multiple servers. Please read the following
> and let us know if you agree.

> Acronyms:
> =========

> ERE	-	Extension Request Extension

> General:
> ========

> We started the discussion by talking about scenarios where multiple IP
> addresses are consumed, and how the protocol would work in such
> environments.

> Can we handle many global and/or site prefixes assigned to a link?
> ==================================================================

> Yes. We need a method by which a client can discover what prefixes are
> available on the link. This is done as follows:

> 	- Sophisticated clients MAY include a ERE in their Solicit which
> 	  requests that servers advertise their prefixes [a].  Servers
> 	  receiving an ERE with the prefix request will advertise
> 	  available prefixes on the client's link by including IP
> 	  address extensions for each prefix, which contains the network
> 	  address and prefix size (subscriber ID is zero).

> 	- Since solicits may generate more than one advertisement
> 	  (multiple servers), a Solicit ID provided in the Solicit (and
> 	  echoed back in the Advertisement) is used to link
> 	  advertisements to their solicitations [b].

> 	- Simple clients do not include the prefix request ERE, and thus
> 	  only care to get *some* set of one or more IP addresses
> 	  returned by the server. They make the number of IP addresses
> 	  known by including empty IP address extensions in their DHCP
> 	  request.

> 	- Relay agents forward traffic to remote DHCP servers. DHCP
> 	  servers have been configured with the mappings from the Relay
> 	  agent's prefix to available prefixes [c].

> 	- A client includes the IP address extensions from the
> 	  advertisement in the Request to signal to the server what
> 	  flavor IP address(es) it would like.


> Requesting additional addresses
> ===============================

> New requests for addresses are differentiated through the use of the XID
> cache implemented on the server. A different XID means a new request.
> The same means a retransmission. We understand that there is a window
> where the XID cache entry may timeout and the server may allocate
> additional IP addresses to the client. We manage that problem by
> specifying that the XID TIMEOUT be 3 hours by default. We suggest
> that somewhat short lifetimes be specified such that IP addresses
> allocated which the client does not know about are reclaimed in a
> timely fashion. If after this draft goes to PS, and we (WG) find that
> this problem proves to be more of a problem than we anticipated, we
> will revisit this issue.

> 3) Router Advertisements

> There has been some traffic on the list wondering whether having a
> single 'M' bit in the header in router advertisements provided enough
> granularity, or whether their should be an 'M' bit in the individual
> prefixes. In other words, there may be some problems in environments
> where stateless and stateful config are combined. Some folks suggested
> that multiple router advertisements would be sent, some for stateless,
> others for stateful.

> We felt that this issue just wasn't a problem, so long as the stateless
> and stateful client implementations operate independently from one
> another.

> 4) Reconfigure/Renumbering

> We talked about whether or not we should have a renumbering section in
> the appendix. (Yes). We also talked about reconfigure, and whether we
> need both schemes or whether we could discard one of them. We've
> identified additional semantics of RECONFIGURE, which we'll present in a
> separate E-mail message.

> 5) Authentication

> We felt that people calling into question whether or not our
> authentication model is adequate given the progress made with DHCPv4
> authentication need to raise specific points regarding potential
> defects in our specification. Our response is that it is up to those
> people to point out the defects to the WG, as we see our model as
> having been validated before. "I'm uncomfortable" isn't a helpful
> comment - "I'm uncomfortable because of a, b, and c" *is*.

> 6) API interface for clients

> I proposed an informational RFC which would describe an API which could
> be used by applications to make requests to and retrieve data from a
> DHCPv6 client. The other authors thought this was a good idea. What do
> you think?

> 7) Multiple Interfaces, multiple choices for application layer extensions.

> Erik raised a concern about how a client would present a coherent view
> of the data returned from multiple interfaces when such data might
> conflict. I suggested we use the method the Solaris DHCPv4 client uses
> to resolve the conflict - the user picks a "primary" interface, and the
> data associated with that interface is authoritative. I've heard no
> comment for or against this proposal. Your input is welcomed.


> Footnotes, etc
> ==============

> 	[a]	ERE needs specification

> 	[b]	The solicit ID needs specification. Perhaps 8 bits out
> 		of the reserved field could be used? Is 8 bits enough? A
> 		solict ID of 0 means that client doesn't match
> 		advertisements to solicits.

> 	[c]	This "mapping" seems rather nasty. I'm wondering if we
> 		should make network administrator's life easier by
> 		having the relay agent enumerate all of the prefixes
> 		available on the client's link, rather than force the
> 		DHCP server administrator to configure this. I think we
> 		need to do this. Comments?



> Mike Carney
> SNT Internet Engineering
> Sun Microsystems, Inc.
> Mailstop UBUR02-212
> One Network Drive
> Burlington, MA 01803



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 11:51:13 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA28315
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 11:51:12 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA06204;
	Thu, 23 Dec 1999 11:42:14 -0500 (EST)
Received: from mailhost.iprg.nokia.com (mailhost.iprg.nokia.com [205.226.5.12])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA27191
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:42:02 -0500 (EST)
Received: from darkstar.iprg.nokia.com (darkstar.iprg.nokia.com [205.226.5.69]) by mailhost.iprg.nokia.com (8.8.8/8.6.10) with ESMTP id IAA19972; Thu, 23 Dec 1999 08:40:59 -0800 (PST)
Received: (from root@localhost)
	by darkstar.iprg.nokia.com (8.9.3/8.9.3-VIRSCAN) id IAA09478;
	Thu, 23 Dec 1999 08:40:59 -0800
X-Virus-Scanned:  Thu, 23 Dec 1999 08:40:59 -0800 Nokia Silicon Valley AntiVirus Appliance
Received: from <charliep@iprg.nokia.com> (charliep.iprg.nokia.com [205.226.2.89]) by darkstar.iprg.nokia.com  SMTP/WTS (12.69)
 xma009394; Thu, 23 Dec 99 08:40:57 -0800
Sender: owner-dhcp-v6@bucknell.edu
Message-ID: <38625099.CD1DF441@iprg.nokia.com>
Date: Thu, 23 Dec 1999 08:40:57 -0800
From: "Charles E. Perkins" <charliep@IPRG.NOKIA.COM>
Organization: Nokia Research Center
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
X-Accept-Language: en
MIME-Version: 1.0
To: <dhcp-v6@bucknell.edu>
CC: dhcp-v6@bucknell.edu
Subject: Re: [Fwd: [Fwd: Re: DHCPv6 issues status]]
References: <199912231601.LAA08881@rotala.raleigh.ibm.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: dhcp-v6@bucknell.edu
X-Sender: charliep@iprg.nokia.com
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN
Content-Transfer-Encoding: 7bit


>>>  - Client wants N addresses (even if the server by default hands out M which
>>>    could be more or less than N)

The scenario was, as I understood it, that the client wants exactly
N addresses at some particular point in time.

>> Client includes an IP address extension.
>> If it gets M (<N) addresses, it asks again.
>> If it gets K (>N) addresses, it releases some.
> 
>> The last case is pretty speculative, and I don't think it should
>> be a common case.  Thus I don't see that we need to devise special
>> protocol to handle it.

I stand by this answer.

> This last case is the most interesting one, and is the one that has
> caused the confusion/complexity so far. This is one of the critical
> aspects of the "request more addresses" model.  Saying it won't be a
> common case and we don't need to worry about it now seems odd. The
> spec needs to nail down the details and semantics of this type of
> operation precisely because they are non-trivial.

What I have always said is that we should allow a DHCP client to get
a new IP address whenever it wants one.  There is no contradiction
whatsoever between this, and my answer to the question that Erik raised.
I do not consider my answer to be "odd".

I also still do not believe that we have to devise special protocol
to handle the case for a client that wants exactly N addresses and
no more and no less.  We can get pretty close with what we have now,
by allowing a client to append N IP address extensions to its DHCP
Request.  I still suggest that most clients will not want to do this,
and that if they did they might just as well append a single IP address
extension and wait to see what comes back from the server.  Or servers!

Even with that long bit of clarification, I stand by my original answer.

Furthermore, I believe that it is not the responsibility of the
DHCP specification to nail down every bit of application-level
programming related to DHCP.  We just have to make the spec work
and solve the needs for stateless address autoconfiguration.  I think
that what you are asking for fits much better in the realm of
policy management for DHCP server database configuration.

> Alternatively, if it really is the case that this won't be a common
> scenario, why should the protocol support it?

The protocol should support the allocation of new IP addresses to
clients
whenever they ask for them.  Perhaps the local administration will
configure their server to be less generous.  Perhaps the local
administration
will configure their server(s) to be very generous and hand out several
IP addresses at every request.

Charlie P.



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 11:51:21 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA28327
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 11:51:20 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA31459;
	Thu, 23 Dec 1999 11:40:55 -0500 (EST)
Received: from fwns2.raleigh.ibm.com (fwns2d.raleigh.ibm.com [204.146.167.236])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA16459
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:40:52 -0500 (EST)
Received: from rtpmail03.raleigh.ibm.com (rtpmail03.raleigh.ibm.com [9.37.172.47])
	by fwns2.raleigh.ibm.com (8.9.0/8.9.0/RTP-FW-1.2) with ESMTP id LAA06150
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:40:50 -0500
Received: from rotala.raleigh.ibm.com (rotala.raleigh.ibm.com [9.37.82.31])
	by rtpmail03.raleigh.ibm.com (8.8.5/8.8.5/RTP-ral-1.1) with ESMTP id LAA30380
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:40:51 -0500
Received: from rotala.raleigh.ibm.com (localhost [127.0.0.1]) by rotala.raleigh.ibm.com (8.9.3/8.7/RTP-ral-1.0) with ESMTP id LAA08987 for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:39:23 -0500
Message-Id: <199912231639.LAA08987@rotala.raleigh.ibm.com>
To: <dhcp-v6@bucknell.edu>
Subject: DHCPv6 authentication issues
In-Reply-To: Message from Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> 
   of "Thu, 16 Dec 1999 11:19:58 EST." <Roam.SIMC.2.0.6.945359149.27374.mwc@atlantic> 
Date: Thu, 23 Dec 1999 11:39:23 -0500
From: Thomas Narten <narten@raleigh.ibm.com>
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

Mike Carney - SNT Internet Engineering <Michael.Carney@East.Sun.COM> writes:

> 5) Authentication

> We felt that people calling into question whether or not our
> authentication model is adequate given the progress made with DHCPv4
> authentication need to raise specific points regarding potential
> defects in our specification. Our response is that it is up to those
> people to point out the defects to the WG, as we see our model as
> having been validated before. "I'm uncomfortable" isn't a helpful
> comment - "I'm uncomfortable because of a, b, and c" *is*.

I'm surprised at this response. The current v4 authentication work is
the result of a lot of effort from folks that really are interested in
making authentication of DHCPv4 work and be deployable in practice.
While I agree that it is desirable for those folks to provide input to
the DHCPv6 effort, this group needs to be proactive in looking at that
work and extracting the relevant parts. Placing the burden on others
to point out what's wrong with the current mechanism is a cop out.

Also, while the current DHCPv6 authentication model may have been
validated in the past, much of the v4 authentication work has taken
place since then. The world has changed, and this effort needs to be
in sync with those changes.

Thomas
 



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 11:56:29 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id LAA28548
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 11:56:29 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id LAA25212;
	Thu, 23 Dec 1999 11:54:35 -0500 (EST)
Received: from mailhost.iprg.nokia.com (mailhost.iprg.nokia.com [205.226.5.12])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id LAA05215
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 11:54:28 -0500 (EST)
Received: from darkstar.iprg.nokia.com (darkstar.iprg.nokia.com [205.226.5.69]) by mailhost.iprg.nokia.com (8.8.8/8.6.10) with ESMTP id IAA21066 for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 08:53:57 -0800 (PST)
Received: (from root@localhost)
	by darkstar.iprg.nokia.com (8.9.3/8.9.3-VIRSCAN) id IAA19306
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 08:53:57 -0800
X-Virus-Scanned:  Thu, 23 Dec 1999 08:53:57 -0800 Nokia Silicon Valley AntiVirus Appliance
Received: from <charliep@iprg.nokia.com> (charliep.iprg.nokia.com [205.226.2.89]) by darkstar.iprg.nokia.com  SMTP/WTS (12.69)
 xma019143; Thu, 23 Dec 99 08:53:52 -0800
Sender: owner-dhcp-v6@bucknell.edu
Message-ID: <386253A0.2B538CB7@iprg.nokia.com>
Date: Thu, 23 Dec 1999 08:53:52 -0800
From: "Charles E. Perkins" <charliep@IPRG.NOKIA.COM>
Organization: Nokia Research Center
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
X-Accept-Language: en
MIME-Version: 1.0
To: <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 and multiple servers
References: <199912231630.LAA08965@rotala.raleigh.ibm.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: dhcp-v6@bucknell.edu
X-Sender: charliep@iprg.nokia.com
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN
Content-Transfer-Encoding: 7bit

Thomas Narten wrote:


> There was another case discussed, where a client has one interface,
> but may want to interact with multiple servers *simultaneously*.  By
> interact, I mean actually get some parameters from one server, other
> parameters from another. Is your position that this second case should
> not be pursued?  (I personally think that not handling this case is
> reasonable, but haven't quite gotten a sense of what the WG thinks
> here.) If this case is to be pursued, however, I believe explicit text
> needs to go into the document explaining when this is desirable, how a
> client decides it needs to interact with multiple servers, etc.

I do not see this.  If a client finds out about a DHCP server,
and sends a DHCP request, and gets a DHCP reply, then that client
has a consistent state with that DHCP server.

If somebody in the future can manage to devise a platform that can
realistically use multiple DHCP servers for a single network interface,
and the existing protocol does not suffer harm, why do we need to make
any specification about whether or not such a future platform does
such a thing?  Why do we need to specify how the client resolves
the answers?  Why do we need to explain the client's motivation?

As an example, if a client today goes out and interacts with multiple
timeservers, it isn't NTP's fault if something goes wrong!

If a client today goes out and dials up two RADIUS servers, it isn't
RADIUS's fault if something goes wrong on the client.

If a client today responds to multiple SNMP masters, it is not SNMP's
fault if something goes wrong.

If a client today gets mail from multiple IMAP servers, it is not IMAP's
fault if the client does the wrong thing.

I could go on and make a similar example out of every other protocol
today, probably...

Furthermore, we can't enforce any restriction on what the client might
do.  We don't have any realistic detection mechanism.  The best remedy
is to have reasonable DHCP server management.  Just as it is the best
remedy in every other server case.


Charlie P.



From owner-dhcp-v6@bucknell.edu  Thu Dec 23 12:11:28 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA28973
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 12:11:27 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA22561;
	Thu, 23 Dec 1999 12:10:34 -0500 (EST)
Received: from mailhost.iprg.nokia.com (mailhost.iprg.nokia.com [205.226.5.12])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id MAA09569
	for <dhcp-v6@bucknell.edu>; Thu, 23 Dec 1999 12:10:31 -0500 (EST)
Received: from darkstar.iprg.nokia.com (darkstar.iprg.nokia.com [205.226.5.69]) by mailhost.iprg.nokia.com (8.8.8/8.6.10) with ESMTP id JAA22253; Thu, 23 Dec 1999 09:09:29 -0800 (PST)
Received: (from root@localhost)
	by darkstar.iprg.nokia.com (8.9.3/8.9.3-VIRSCAN) id JAA27862;
	Thu, 23 Dec 1999 09:09:29 -0800
X-Virus-Scanned:  Thu, 23 Dec 1999 09:09:29 -0800 Nokia Silicon Valley AntiVirus Appliance
Received: from <charliep@iprg.nokia.com> (charliep.iprg.nokia.com [205.226.2.89]) by darkstar.iprg.nokia.com  SMTP/WTS (12.69)
 xma027779; Thu, 23 Dec 99 09:09:27 -0800
Sender: owner-dhcp-v6@bucknell.edu
Message-ID: <38625747.75B0F68F@iprg.nokia.com>
Date: Thu, 23 Dec 1999 09:09:27 -0800
From: "Charles E. Perkins" <charliep@IPRG.NOKIA.COM>
Organization: Nokia Research Center
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
X-Accept-Language: en
MIME-Version: 1.0
To: <dhcp-v6@bucknell.edu>
CC: dhcp-v6@bucknell.edu
Subject: Re: DHCPv6 authentication issues
References: <199912231639.LAA08987@rotala.raleigh.ibm.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: dhcp-v6@bucknell.edu
X-Sender: charliep@iprg.nokia.com
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN
Content-Transfer-Encoding: 7bit

Thomas Narten wrote:

> I'm surprised at this response.  ...   Placing the burden on others
> to point out what's wrong with the current mechanism is a cop out.

The current model for DHCPv6 validation has been through the ringer
for years in the context of Mobile IP.  No one has yet suggested that
it has significant flaws, at least not in any forum that I know of.
It may not be perfect, and it doesn't solve key distribution.

Specifying and adapting the use of a known good mechanism is not a
cop out, and I don't think you should portray it that way.  Given that
we have done a good job of specifying and adapting, then for us to ask
for specific comments is not a cop-out either.

Having said that, I do not disagree that studying the DHCPv4
authentication
mechanism would likely be worthwhile.  I just disagree with your cop-out
accusation.  Furthermore, it would be even more worthwhile to do further
research into authentication mechanisms.  It's all a matter of time and
effort.  I thought we had made a pretty good effort, short of trying
to hit a moving target.

Do you have a specific demand in mind from your perspective as AD?
Or, do you have a specific requirement in mind in your role as a
possible implementer of the protocol?


> Also, while the current DHCPv6 authentication model may have been
> validated in the past, much of the v4 authentication work has taken
> place since then. The world has changed, and this effort needs to be
> in sync with those changes.

Can you mention a specific change that we should focus our awareness on?

Charlie P.



From owner-dhcp-v4@bucknell.edu  Thu Dec 23 15:19:26 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA02472
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 15:19:25 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA14995;
	Thu, 23 Dec 1999 15:17:48 -0500 (EST)
Received: from ariel.gi.com (ariel.gi.com [168.84.84.10])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id PAA30486
	for <dhcp-v4@bucknell.edu>; Thu, 23 Dec 1999 15:17:22 -0500 (EST)
Received: from ntas0028.gi.com ([168.84.84.98]) by GI.COM (PMDF V5.2-31 #38811)
 with ESMTP id <01JJUA1AGDXKCDSSTI@GI.COM> for dhcp-v4@bucknell.edu; Thu,
 23 Dec 1999 12:17:29 PST
Received: by ntas0028.gi.com with Internet Mail Service (5.5.2650.21)
	id <ZNHKKXFN>; Thu, 23 Dec 1999 12:20:46 -0500
Content-return: allowed
Date: Thu, 23 Dec 1999 15:17:24 -0500
From: "Lalwaney, Poornima (SD-EX)" <PLalwaney@gi.com>
Subject: DHCP Continuation Option
To: <dhcp-v4@bucknell.edu>
Message-id: <97DEDE66B3DCD11199D200805FA71BE2021D5FE5@ntas0027.gi.com>
MIME-version: 1.0
X-Mailer: Internet Mail Service (5.5.2650.21)
Content-type: multipart/alternative;
	boundary="----_=_NextPart_001_01BF4D6A.0C27C6CE"
Reply-To: PLalwaney@gi.com
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.

------_=_NextPart_001_01BF4D6A.0C27C6CE
Content-Type: text/plain


Is there a later version of the draft  <draft-ietf-dhc-options-cont-00.txt>
by Bill Arbaugh and Angelos Keromytis on DHCP continuation option code.
This draft expired in May 98 but I saw some reference to this a couple of
months ago when  the authentication draft was being discussed on this list. 

Are there other approaches besides the continuation option to accommodate an
option size > 255 bytes?


Thanks..
Poornima

------_=_NextPart_001_01BF4D6A.0C27C6CE
Content-Type: text/html
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Dus-ascii">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
5.5.2650.12">
<TITLE>DHCP Continuation Option</TITLE>
</HEAD>
<BODY>
<BR>

<P><FONT FACE=3D"Times New Roman">Is there a later version of the =
draft&nbsp; &lt;draft-ietf-dhc-options-cont-00.txt&gt; by Bill Arbaugh =
and Angelos Keromytis on DHCP continuation option code.&nbsp; This =
draft expired in May 98 but I saw some reference to this a couple of =
months ago when&nbsp; the authentication draft was being discussed on =
this list. </FONT></P>

<P><FONT FACE=3D"Times New Roman">Are there other approaches besides =
the continuation option to accommodate an option size &gt; 255 =
bytes?</FONT>
</P>
<BR>

<P><FONT FACE=3D"Times New Roman">Thanks..</FONT>
<BR><FONT FACE=3D"Times New Roman">Poornima</FONT>
</P>

</BODY>
</HTML>
------_=_NextPart_001_01BF4D6A.0C27C6CE--



From owner-dhcp-v4@bucknell.edu  Thu Dec 23 16:16:51 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA03210
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 16:16:51 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA06959;
	Thu, 23 Dec 1999 16:15:34 -0500 (EST)
Received: from codex.cis.upenn.edu (CODEX.CIS.UPENN.EDU [158.130.6.15])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id QAA04959
	for <dhcp-v4@bucknell.edu>; Thu, 23 Dec 1999 16:15:24 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by codex.cis.upenn.edu (8.8.5/8.8.5) with SMTP id QAA18842;
	Thu, 23 Dec 1999 16:15:20 -0500 (EST)
Date: Thu, 23 Dec 1999 16:15:20 -0500 (EST)
From: "William A. Arbaugh" <waa@dsl.cis.upenn.edu>
To: <dhcp-v4@bucknell.edu>
cc: dhcp-v4@bucknell.edu
Subject: Re: DHCP Continuation Option
In-Reply-To: <97DEDE66B3DCD11199D200805FA71BE2021D5FE5@ntas0027.gi.com>
Message-ID: <Pine.GSO.3.95.991223161334.18821A-100000@codex.cis.upenn.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: waa@dsl.cis.upenn.edu
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


We're going to re-issue since Gupta's draft requires (as likely will any
Public Key solution) it.  I just haven't gotten to it yet.  Perhaps next
week.

Bill


On Thu, 23 Dec 1999, Lalwaney, Poornima (SD-EX) wrote:

> 
> Is there a later version of the draft  <draft-ietf-dhc-options-cont-00.txt>
> by Bill Arbaugh and Angelos Keromytis on DHCP continuation option code.
> This draft expired in May 98 but I saw some reference to this a couple of
> months ago when  the authentication draft was being discussed on this list. 
> 
> Are there other approaches besides the continuation option to accommodate an
> option size > 255 bytes?
> 
> 
> Thanks..
> Poornima
> 



From owner-dhcp-v4@bucknell.edu  Thu Dec 23 16:17:23 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id QAA03222
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 23 Dec 1999 16:17:23 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id QAA14871;
	Thu, 23 Dec 1999 16:17:15 -0500 (EST)
Received: from codex.cis.upenn.edu (CODEX.CIS.UPENN.EDU [158.130.6.15])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id QAA00785
	for <dhcp-v4@bucknell.edu>; Thu, 23 Dec 1999 16:17:07 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by codex.cis.upenn.edu (8.8.5/8.8.5) with SMTP id QAA18873
	for <dhcp-v4@bucknell.edu>; Thu, 23 Dec 1999 16:17:06 -0500 (EST)
Date: Thu, 23 Dec 1999 16:17:06 -0500 (EST)
From: "William A. Arbaugh" <waa@dsl.cis.upenn.edu>
To: <dhcp-v4@bucknell.edu>
Subject: Don't forget those authentication draft edits!
Message-ID: <Pine.GSO.3.95.991223161550.18821B-100000@codex.cis.upenn.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: waa@dsl.cis.upenn.edu
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


I know some people still have comments that they haven't sent me yet.  I
hope to re-issue the draft next week.  So please .....

Bill

-----------------------------------------------------------------------
Bill Arbaugh			   
email:  waa@dsl.cis.upenn.edu 
web:    http://www.cis.upenn.edu/~waa
-----------------------------------------------------------------------  



From owner-dhcp-v6@bucknell.edu  Mon Dec 27 15:09:40 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id PAA07802
	for <DHC-ARCHIVE@odin.ietf.org>; Mon, 27 Dec 1999 15:09:39 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id PAA15252;
	Mon, 27 Dec 1999 15:02:14 -0500 (EST)
Received: from roll.mentat.com (mentat.com [192.88.122.129])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id PAA18488
	for <dhcp-v6@bucknell.edu>; Mon, 27 Dec 1999 15:02:03 -0500 (EST)
Received: from leo.mentat.com (leo [192.88.122.132])
	by roll.mentat.com (8.9.1b+Sun/8.9.1) with ESMTP id MAA00232
	for <dhcp-v6@bucknell.edu>; Mon, 27 Dec 1999 12:01:44 -0800 (PST)
Received: from feller.mentat.com (feller [192.88.122.144])
	by leo.mentat.com (8.9.1b+Sun/8.9.1) with ESMTP id MAA27890
	for <dhcp-v6@bucknell.edu>; Mon, 27 Dec 1999 12:01:45 -0800 (PST)
Received: (from tim@localhost)
	by feller.mentat.com (8.9.1b+Sun/8.9.1) id MAA00903
	for dhcp-v6@bucknell.edu; Mon, 27 Dec 1999 12:03:30 -0800 (PST)
Date: Mon, 27 Dec 1999 12:03:30 -0800 (PST)
From: Tim Hartrick <tim@mentat.com>
Message-Id: <199912272003.MAA00903@feller.mentat.com>
To: <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 and multiple servers
X-Sun-Charset: US-ASCII
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN





Thomas,

> 
> There was another case discussed, where a client has one interface,
> but may want to interact with multiple servers *simultaneously*.  By
> interact, I mean actually get some parameters from one server, other
> parameters from another. Is your position that this second case should
> not be pursued?  (I personally think that not handling this case is
> reasonable, but haven't quite gotten a sense of what the WG thinks
> here.) If this case is to be pursued, however, I believe explicit text
> needs to go into the document explaining when this is desirable, how a 
> client decides it needs to interact with multiple servers, etc.
>

It depends on what you mean by "not handling this case".  If you mean
that we should make it impossible for a sufficiently intelligent client
to handle this case, and put text in the document which says you MUST NOT
implement such a client, I disagree.  If you mean that we don't need
to specify the details of such an interaction now, other than making sure
that the specification does not prohibit the possibility, then I agree.

My only point in bringing this up previously was that it seemed obvious
to me that mobile nodes may well need to be able to "interact" with multiple
DHCPv6 servers in the home network and the remote network.  If we
put something into DHCPv6 that prevents such clients from ever existing
we are making a mistake.  Currently, I don't see any thing that would pre-
vent such a client from existing.  I just want to make sure that nothing
gets added that would prevent such a client from existing.

> 
> > 7) Multiple Interfaces, multiple choices for application layer extensions.
> 
> > Erik raised a concern about how a client would present a coherent view
> > of the data returned from multiple interfaces when such data might
> > conflict. I suggested we use the method the Solaris DHCPv4 client uses
> > to resolve the conflict - the user picks a "primary" interface, and the
> > data associated with that interface is authoritative. I've heard no
> > comment for or against this proposal. Your input is welcomed.
> 
 
It seems that that we could do a similar thing for handling multiple
servers per interface for the mobility case.  One could have the notion of
a home DHCPv6 server and a remote DHCPv6 server and designate which
parameters are authoritative from which server.

This does not need to specified now to the Nth detail now.  We just need to
make sure that it is not prohibited in the base specification by engineering
or politics.



tim



From owner-dhcp-v6@bucknell.edu  Wed Dec 29 13:10:46 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id NAA01178
	for <DHC-ARCHIVE@odin.ietf.org>; Wed, 29 Dec 1999 13:10:45 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id NAA21787;
	Wed, 29 Dec 1999 13:05:47 -0500 (EST)
Received: from hygro.adsl.duke.edu (hygro.adsl.duke.edu [152.16.64.159])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id NAA08543
	for <dhcp-v6@bucknell.edu>; Wed, 29 Dec 1999 13:05:33 -0500 (EST)
Received: from hygro.adsl.duke.edu (localhost [127.0.0.1])
	by hygro.adsl.duke.edu (8.8.7/8.8.7) with ESMTP id NAA13564
	for <dhcp-v6@bucknell.edu>; Wed, 29 Dec 1999 13:04:13 -0500
Message-Id: <199912291804.NAA13564@hygro.adsl.duke.edu>
To: <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 and multiple servers 
In-Reply-To: Message from Tim Hartrick <tim@mentat.com> 
   of "Mon, 27 Dec 1999 12:03:30 PST." <199912272003.MAA00903@feller.mentat.com> 
Date: Wed, 29 Dec 1999 13:04:13 -0500
From: Thomas Narten <narten@raleigh.ibm.com>
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

Tim,

> > There was another case discussed, where a client has one interface,
> > but may want to interact with multiple servers *simultaneously*.  By
> > interact, I mean actually get some parameters from one server, other
> > parameters from another. Is your position that this second case should
> > not be pursued?  (I personally think that not handling this case is
> > reasonable, but haven't quite gotten a sense of what the WG thinks
> > here.) If this case is to be pursued, however, I believe explicit text
> > needs to go into the document explaining when this is desirable, how a 
> > client decides it needs to interact with multiple servers, etc.
> >

> It depends on what you mean by "not handling this case".  If you mean
> that we should make it impossible for a sufficiently intelligent client
> to handle this case, and put text in the document which says you MUST NOT
> implement such a client, I disagree.  If you mean that we don't need
> to specify the details of such an interaction now, other than making sure
> that the specification does not prohibit the possibility, then I
> agree.

I did not intend to imply that there should be MUST NOT language. I
certainly don't believe that.

What I would like to be clear on (and maybe this is already clear to
everyone else) is whether it is requirement that clients are expected
to make use of multiple servers, in the simultaneous sense. Although
the spec doesn't disallow it, it also doesn't suggest that clients do
this. If there is an expectation that clients do need to handle this
case, more text in the document is needed. If clients do not need to
do this today, than the current document is fine in this regard.

> My only point in bringing this up previously was that it seemed obvious
> to me that mobile nodes may well need to be able to "interact" with multiple
> DHCPv6 servers in the home network and the remote network.  If we
> put something into DHCPv6 that prevents such clients from ever existing
> we are making a mistake.  Currently, I don't see any thing that would pre-
> vent such a client from existing.  I just want to make sure that nothing
> gets added that would prevent such a client from existing.

I totally agree.

> > 
> > > 7) Multiple Interfaces, multiple choices for application layer extensions.
> > 
> > > Erik raised a concern about how a client would present a coherent view
> > > of the data returned from multiple interfaces when such data might
> > > conflict. I suggested we use the method the Solaris DHCPv4 client uses
> > > to resolve the conflict - the user picks a "primary" interface, and the
> > > data associated with that interface is authoritative. I've heard no
> > > comment for or against this proposal. Your input is welcomed.
> > 
>  
> It seems that that we could do a similar thing for handling multiple
> servers per interface for the mobility case.  One could have the notion of
> a home DHCPv6 server and a remote DHCPv6 server and designate which
> parameters are authoritative from which server.

I'm not sure this is so clear. I can imagine that there are cases
where the DHCP-learned parameter one uses (i.e., DNS server) depends
on whether the node is pretending to be mobile (i.e., using its home
address) or whether it is acting directly attached (i.e., using its
temporary address directly). So this case may need more thought.

> This does not need to specified now to the Nth detail now.  We just need to
> make sure that it is not prohibited in the base specification by engineering
> or politics.

Agreed.

Thomas



From owner-dhcp-v4@bucknell.edu  Wed Dec 29 17:40:31 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id RAA04850
	for <DHC-ARCHIVE@odin.ietf.org>; Wed, 29 Dec 1999 17:40:28 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id RAA06237;
	Wed, 29 Dec 1999 17:38:51 -0500 (EST)
Received: from toccata.fugue.com (toccata.fugue.com [204.152.188.66])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id RAA12481
	for <dhcp-v4@bucknell.edu>; Wed, 29 Dec 1999 17:38:28 -0500 (EST)
Received: from grosse.manhattan.fugue.com (tundruk.manhattan.fugue.com [160.79.19.71]) by toccata.fugue.com (8.9.3/8.6.11) with ESMTP id OAA11419; Wed, 29 Dec 1999 14:33:13 -0800 (PST)
Received: from grosse.manhattan.fugue.com (localhost [127.0.0.1]) by grosse.manhattan.fugue.com (8.9.3/8.6.11) with ESMTP id RAA08840; Wed, 29 Dec 1999 17:38:24 -0500 (EST)
Message-Id: <199912292238.RAA08840@grosse.manhattan.fugue.com>
To: <dhcp-v4@bucknell.edu>
cc: dhcp-v4@bucknell.edu
Subject: Re: Vote for subnet mask relay agent sub-option 
In-Reply-To: Message from "Michael W. Patrick" <mpatrick@noah.dma.isg.mot.com> 
   of "Wed, 29 Dec 1999 17:09:38 EST." <199912292209.RAA15273@noah.dma.isg.mot.com> 
Date: Wed, 29 Dec 1999 17:38:23 -0500
From: Ted Lemon <mellon@isc.org>
Reply-To: mellon@isc.org
Sender: owner-dhcp-v4@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN


Consensus means nobody disagrees, not that a majority agrees.   So
calling for a vote is inappropriate, and if you get a majority voting
in favour of including your subnet mask suboption in the relay agent
information option draft, this will not mean that you have a working
group consensus to proceed with the draft.

To describe what is going on as a dispute seems unnecessary - it
implies dischord where none exists.  I do not think that the subnet
mask suboption is necessarily a bad idea.  I think it belongs in a
draft because it implies a rather impressive expansion of what a DHCP
server does.  I think that two paragraphs describing in minimal detail
how a relay agent should specify this option is not sufficient - the
behaviour of the server also needs to be described.

In particular, what you have now doesn't discuss:

    - Whether this suboption can be used when a relay agent is
      configured to relay to more than one DHCP server
    - If more than one server is configured, how these servers will
      do address allocation in such a way that they do not step on
      each other's allocation pools
    - What happens when a router that's configured to support a
      particular subnet disappears, along with the subnet it supported
    - Discussion as to how network administrators should configure
      routers that support this option so as to avoid configuration
      errors such as two routers serving two physical networks that
      choose the same subnet
    - Justification for why this option is a better choice than just
      configuring the DHCP server with information about the subnet
    - Suggestions or recommendations for what configuration parameters
      should be available on a DHCP server so that it can
      appropriately serve one of these ad-hoc subnets

My reason for suggesting that this be in a seperate draft is that we
have waited a long time for the Relay Agent Information Option draft
to move to proposed standard.  It seems to me that the subnet mask
draft is not trivial, and needs to be discussed further.   Holding
back the RAIO draft because we need to flesh out the subnet mask
suboption strikes me as a bad move.   Rushing the subnet mask
suboption and not fleshing out its description likewise seems like a
bad move.

In my mind, this leaves only one choice, which is the one I suggested:
split the subnet mask option out into its own draft.  If I have
appeared to be contentious in expressing this point of view, I
apologize.

			       _MelloN_



From owner-dhcp-v6@bucknell.edu  Thu Dec 30 01:51:58 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id BAA12817
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 30 Dec 1999 01:51:57 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id BAA03665;
	Thu, 30 Dec 1999 01:48:01 -0500 (EST)
Received: from roll.mentat.com (mentat.com [192.88.122.129])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id BAA02302
	for <dhcp-v6@bucknell.edu>; Thu, 30 Dec 1999 01:47:55 -0500 (EST)
Received: from leo.mentat.com (leo [192.88.122.132])
	by roll.mentat.com (8.9.1b+Sun/8.9.1) with ESMTP id WAA09058
	for <dhcp-v6@bucknell.edu>; Wed, 29 Dec 1999 22:47:40 -0800 (PST)
Received: from feller.mentat.com (feller [192.88.122.144])
	by leo.mentat.com (8.9.1b+Sun/8.9.1) with ESMTP id WAA21287
	for <dhcp-v6@bucknell.edu>; Wed, 29 Dec 1999 22:47:40 -0800 (PST)
Received: (from tim@localhost)
	by feller.mentat.com (8.9.1b+Sun/8.9.1) id WAA01766
	for dhcp-v6@bucknell.edu; Wed, 29 Dec 1999 22:49:27 -0800 (PST)
Date: Wed, 29 Dec 1999 22:49:27 -0800 (PST)
From: Tim Hartrick <tim@mentat.com>
Message-Id: <199912300649.WAA01766@feller.mentat.com>
To: <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 and multiple servers
X-Sun-Charset: US-ASCII
Reply-To: dhcp-v6@bucknell.edu
Sender: owner-dhcp-v6@bucknell.edu
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN




Thomas,

> 
> What I would like to be clear on (and maybe this is already clear to
> everyone else) is whether it is requirement that clients are expected
> to make use of multiple servers, in the simultaneous sense. Although
> the spec doesn't disallow it, it also doesn't suggest that clients do
> this. If there is an expectation that clients do need to handle this
> case, more text in the document is needed. If clients do not need to
> do this today, than the current document is fine in this regard.
>

There should be no MUST language either.  It should be possible for a
client to exist which will only interact with one server even if multiple
servers are present.

> >  
> > It seems that that we could do a similar thing for handling multiple
> > servers per interface for the mobility case.  One could have the notion of
> > a home DHCPv6 server and a remote DHCPv6 server and designate which
> > parameters are authoritative from which server.
> 
> I'm not sure this is so clear. I can imagine that there are cases
> where the DHCP-learned parameter one uses (i.e., DNS server) depends
> on whether the node is pretending to be mobile (i.e., using its home
> address) or whether it is acting directly attached (i.e., using its
> temporary address directly). So this case may need more thought.
> 

Yes, it is complex.  It really requires that there be some linkage between
the address being used as source address for communication and the parameters
received from the DHCPv6 server which issued the address being used as the
source address.  Pretty ugly.


tim



From owner-dhcp-v6@bucknell.edu  Thu Dec 30 12:14:33 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA00501
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 30 Dec 1999 12:14:32 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA10542;
	Thu, 30 Dec 1999 12:10:36 -0500 (EST)
Received: from concorde.inria.fr (concorde.inria.fr [192.93.2.39])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id MAA31760
	for <dhcp-v6@bucknell.edu>; Thu, 30 Dec 1999 12:10:29 -0500 (EST)
Received: from givry.inria.fr (givry.inria.fr [193.51.193.144])
	by concorde.inria.fr (8.8.7/8.8.7) with ESMTP id SAA18504
	for <dhcp-v6@bucknell.edu>; Thu, 30 Dec 1999 18:10:24 +0100 (MET)
Received: from givry.inria.fr (givry.inria.fr [193.51.193.144]) by givry.inria.fr (8.7.6/8.7.3) with ESMTP id SAA28738 for <dhcp-v6@bucknell.edu>; Thu, 30 Dec 1999 18:10:24 +0100 (MET)
Message-Id: <199912301710.SAA28738@givry.inria.fr>
From: Francis Dupont <Francis.Dupont@inria.fr>
To: <dhcp-v6@bucknell.edu>
Subject: Re: DHCPv6 address allocation model 
In-reply-to: Your message of Thu, 23 Dec 1999 11:14:48 EST.
             <199912231614.LAA08919@rotala.raleigh.ibm.com> 
Date: Thu, 30 Dec 1999 18:10:23 +0100
Sender: owner-dhcp-v6@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
X-Sender: Francis.Dupont@inria.fr
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

I believe if we split client addresses in :
 - primary addresses, one per prefix, typically a primary address
   is the address a client could get when it uses stateless autoconf.
   Primary addresses are allocated by DHCPv6 servers (one set per client),
   common clients need only primary addresses.
 - secondary addresses which are explicitely needed by some special clients,
   they are extra-addresses (ie. the client has already a primary address for
   this prefix).
   My proposal: servers don't allocate secondary addresses, when a client
   wants a secondary address, it builds it and send a request to the server
   with a filled address extension in order to check if this address is
   available. All management actions should be initiated by the client
   (this removes most of problems of the DHCPv6 allocation model).

Happy new year

Francis.Dupont@inria.fr



From owner-dhcp-v6@bucknell.edu  Thu Dec 30 12:32:37 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id MAA01297
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 30 Dec 1999 12:32:36 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id MAA16980;
	Thu, 30 Dec 1999 12:31:52 -0500 (EST)
Received: from mailhost.iprg.nokia.com (mailhost.iprg.nokia.com [205.226.5.12])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id MAA15114
	for <dhcp-v6@bucknell.edu>; Thu, 30 Dec 1999 12:31:47 -0500 (EST)
Received: from darkstar.iprg.nokia.com (darkstar.iprg.nokia.com [205.226.5.69]) by mailhost.iprg.nokia.com (8.8.8/8.6.10) with ESMTP id JAA24222; Thu, 30 Dec 1999 09:31:14 -0800 (PST)
Received: (from root@localhost)
	by darkstar.iprg.nokia.com (8.9.3/8.9.3-VIRSCAN) id JAA10119;
	Thu, 30 Dec 1999 09:31:14 -0800
X-Virus-Scanned:  Thu, 30 Dec 1999 09:31:14 -0800 Nokia Silicon Valley AntiVirus Appliance
Received: from <charliep@iprg.nokia.com> (charliep.iprg.nokia.com [205.226.2.89]) by darkstar.iprg.nokia.com  SMTP/WTS (12.69)
 xma010035; Thu, 30 Dec 99 09:31:10 -0800
Sender: owner-dhcp-v6@bucknell.edu
Message-ID: <386B96DE.50491E71@iprg.nokia.com>
Date: Thu, 30 Dec 1999 09:31:10 -0800
From: "Charles E. Perkins" <charliep@IPRG.NOKIA.COM>
Organization: Nokia Research Center
X-Mailer: Mozilla 4.5 [en] (X11; I; FreeBSD 2.2.6-RELEASE i386)
X-Accept-Language: en
MIME-Version: 1.0
To: <dhcp-v6@bucknell.edu>
CC: dhcp-v6@bucknell.edu
Subject: Re: DHCPv6 address allocation model
References: <199912301710.SAA28738@givry.inria.fr>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: dhcp-v6@bucknell.edu
X-Sender: charliep@iprg.nokia.com
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN
Content-Transfer-Encoding: 7bit



Hello Francis,

I understand the desire for simplicity in your proposal.
And, I also believe it is a reasonable policy for use by
typical system administrators.

However, I have generally thought that it was the job of
the protocol specification to create the mechanism, while
not constraining the policy.

Thus, I suggest that your policy be followed in the typical
case, but that the protocol specification allow for other
policies.  For instance, a network administrator may desire
to specify other address allocation strategies for particular
nodes, or for all nodes of a particular class, and so on.

Regards,
Charlie P.




Francis Dupont wrote:
> 
> I believe if we split client addresses in :
>  - primary addresses, one per prefix, typically a primary address
>    is the address a client could get when it uses stateless autoconf.
>    Primary addresses are allocated by DHCPv6 servers (one set per client),
>    common clients need only primary addresses.
>  - secondary addresses which are explicitely needed by some special clients,
>    they are extra-addresses (ie. the client has already a primary address for
>    this prefix).
>    My proposal: servers don't allocate secondary addresses, when a client
>    wants a secondary address, it builds it and send a request to the server
>    with a filled address extension in order to check if this address is
>    available. All management actions should be initiated by the client
>    (this removes most of problems of the DHCPv6 allocation model).
> 
> Happy new year
> 
> Francis.Dupont@inria.fr



From owner-dhcp-v6@bucknell.edu  Thu Dec 30 14:31:49 1999
Received: from mail.bucknell.edu (marge.bucknell.edu [134.82.9.1])
	by ietf.org (8.9.1a/8.9.1a) with ESMTP id OAA03696
	for <DHC-ARCHIVE@odin.ietf.org>; Thu, 30 Dec 1999 14:31:49 -0500 (EST)
Received: from localhost (localhost [127.0.0.1])
	by mail.bucknell.edu (8.9.3/8.9.3) with SMTP id OAA14904;
	Thu, 30 Dec 1999 14:27:53 -0500 (EST)
Received: from concorde.inria.fr (concorde.inria.fr [192.93.2.39])
	by mail.bucknell.edu (8.9.3/8.9.3) with ESMTP id OAA04922
	for <dhcp-v6@bucknell.edu>; Thu, 30 Dec 1999 14:27:40 -0500 (EST)
Received: from givry.inria.fr (givry.inria.fr [193.51.193.144])
	by concorde.inria.fr (8.8.7/8.8.7) with ESMTP id UAA23842;
	Thu, 30 Dec 1999 20:27:36 +0100 (MET)
Received: from givry.inria.fr (givry.inria.fr [193.51.193.144]) by givry.inria.fr (8.7.6/8.7.3) with ESMTP id UAA19417; Thu, 30 Dec 1999 20:27:34 +0100 (MET)
Message-Id: <199912301927.UAA19417@givry.inria.fr>
From: Francis Dupont <Francis.Dupont@inria.fr>
To: <dhcp-v6@bucknell.edu>
cc: dhcp-v6@bucknell.edu
Subject: Re: DHCPv6 address allocation model 
In-reply-to: Your message of Thu, 30 Dec 1999 09:31:10 PST.
             <386B96DE.50491E71@iprg.nokia.com> 
Date: Thu, 30 Dec 1999 20:27:33 +0100
Sender: owner-dhcp-v6@bucknell.edu
Reply-To: dhcp-v6@bucknell.edu
X-Sender: Francis.Dupont@inria.fr
X-Listprocessor-Version: 8.2.10/991025/16:55 -- ListProc(tm) by CREN

 In your previous mail you wrote:

   However, I have generally thought that it was the job of
   the protocol specification to create the mechanism, while
   not constraining the policy.
   
   Thus, I suggest that your policy be followed in the typical
   case, but that the protocol specification allow for other
   policies.  For instance, a network administrator may desire
   to specify other address allocation strategies for particular
   nodes, or for all nodes of a particular class, and so on.
   
=> I agree. In a previous message I proposed to put this kind of things
in an application document like we have for routing protocols (by the
way we need a MIB document too). This has many advantages and can
make DHCPv6 specs easier to finish because documents of this kind are
informational only (in the case this will go in a separate document,
I think this is the best solution because current I-Ds are large enough).

Thanks

Francis.Dupont@inria.fr



