From ftp-wg-owner@hethmon.com  Tue Jan 12 16:13:08 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id QAA24527
	for <ftpext-archive@lists.ietf.org>; Tue, 12 Jan 1999 16:13:07 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Tue, 12 Jan 1999 16:04:00 -0500
Received: from rbhub101.chamb.disa.mil (rbhub101.chamb.disa.mil [209.22.120.24] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Tue, 12 Jan 1999 16:03:48 -0500
Received: by rbhub101.chamb.disa.mil with Internet Mail Service (5.5.2232.9)
	id <CW4TDVNJ>; Tue, 12 Jan 1999 16:02:38 -0500
Message-ID: <43B821CCD144D211AB0500204804EE4A2BB21C@rbmail102.chamb.disa.mil>
MIME-Version: 1.0
X-Mailer: Internet Mail Service (5.5.2232.9)
Content-Type: text/plain;
	charset="iso-8859-1"
Date: Tue, 12 Jan 1999 16:03:49 -0500
X-OldDate:   Tue, 12 Jan 1999 16:02:53 -0500
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: "Curtin, William" <curtinw@ftm.disa.mil>
To: "'ftp-wg@hethmon.com'" <ftp-wg@hethmon.com>
Subject: Ftp-WG:   Lang command

I've been trying to think through the addition of a "LANG" command and
associated FEAT response and OPTS request, for the I18N draft (as
recommended from the last call).

Here is my plan:

1. Create a LANG command with parameters associated with the primary
language tag defined in RFC 1766 (i.e. ISO 639 tags). 

2. Recommend that clients, prior to authentication, issue a FEAT command to
discover what languages the server can support.

3. State that servers must support "Welcome banners" and command
responses/errors in a given language to claim support for that language. The
draft would of course state that FTP commands included in responses (e.g.
USER, PORT, etc.) should not be translated into the language being
supported. The server must wait until after receiving a LANG command before
issuing the "welcome message".

4. Require the client issue OPTS for the LANG command with the requested
language prior to authentication.

5. If the server cannot support the language being requested by the client,
it must use the default language. The default language (i-default) is
defined in 1766. Now RFC 2277 states that the default language must be
understandable by english speaking people. Will the recommendation that
servers must support "en" as a default start a flame war?


This will all work fine for new clients and servers but I'm having some
problems figuring out when the welcome message can be sent.

For new clients and servers: MLST recommends the server wait until after a
HOST command is issued to send the "welcome". What if a HOST command is not
sent? Should the I18N draft require MLST be supported and therefore HOST
will be supported (presumably after the LANG command and before the USER
command). Also some discussion has recommended defining a new command for
displaying the "welcome" message. Should a command be defined to do this and
relieve the dependence on HOST? That latter approach seems cleaner to me.
The requirement would be that the client issue a new "MESS" command just
prior to authentication.

For old clients and new servers: How should the server recognize that it is
talking to an old client and that it should probably issue the "Welcome"
after the TCP connection?? If it waits for the LANG, HOST, MESS commands to
first send the "welcome" message it will receive the USER command and the
human associated with the user-PI may miss login information from the
server.

For new clients and old servers: It probably doesn't make a difference as
the old servers will not recognize the LANG command and the client get
responses and messages in the language supported by the server. No
difference than what is happening today.


Guidance from the group would be greatly appreciated.


							bill



From ftp-wg-owner@hethmon.com  Tue Jan 12 17:56:34 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id RAA26408
	for <ftpext-archive@lists.ietf.org>; Tue, 12 Jan 1999 17:56:33 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Tue, 12 Jan 1999 17:53:04 -0500
Received: from phobos.illtel.denver.co.us (dsl-206.169.4.82.wenet.com [206.169.4.82] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Tue, 12 Jan 1999 17:52:52 -0500
Received: from localhost (abelits@localhost) by phobos.illtel.denver.co.us (8.9.1a/8.6.9) with SMTP id OAA09829 for <ftp-wg@hethmon.com>; Tue, 12 Jan 1999 14:50:23 -0800
In-Reply-To: <43B821CCD144D211AB0500204804EE4A2BB21C@rbmail102.chamb.disa.mil>
Message-ID: <Pine.LNX.4.02A.9901121307080.30532-100000@phobos.illtel.denver.co.us>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Date: Tue, 12 Jan 1999 17:52:53 -0500
X-OldDate:   Tue, 12 Jan 1999 14:50:23 -0800 (PST)
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Alex Belits <abelits@phobos.illtel.denver.co.us>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   Lang command

On Tue, 12 Jan 1999, Curtin, William wrote:

> 1. Create a LANG command with parameters associated with the primary
> language tag defined in RFC 1766 (i.e. ISO 639 tags). 

  While I don't think that such functionality is important, if it will be
considered to be useful, I have several comments about implementation.

> 2. Recommend that clients, prior to authentication, issue a FEAT command to
> discover what languages the server can support.
> 
> 3. State that servers must support "Welcome banners" and command
> responses/errors in a given language to claim support for that language. The
> draft would of course state that FTP commands included in responses (e.g.
> USER, PORT, etc.) should not be translated into the language being
> supported.

  Text part of the response to FTP commands in most of cases is not
supposed to be parsed by the client that uses numerical part only, and in
existing implementations text part sometimes is translated into the
locally configured language (say, AIX ftpd). While this practice is often
counterproductive (remote human user may be unable to read the
local language), it's harmless.

> The server must wait until after receiving a LANG command before
> issuing the "welcome message".
>
> 4. Require the client issue OPTS for the LANG command with the requested
> language prior to authentication.
> 
> 5. If the server cannot support the language being requested by the client,
> it must use the default language. The default language (i-default) is
> defined in 1766. Now RFC 2277 states that the default language must be
> understandable by english speaking people. Will the recommendation that
> servers must support "en" as a default start a flame war?

  Unlikely. However it may be necessary to send a language/charset pair
instead of language only -- some languages have multiple charsets, and
even though some people propose the total ban on all charsets other than
Unicode and on all encodings other than UTF-8 (or at lease cripple the
charsets support to make the use of anything but UTF-8 as much painful as
possible), users of other charsets strongly disagree with it. However if
multiple charsets and languages are supported, there must be no
requirement that all servers must support all charsets of the same
language if even only one charset for this language is supported --
computer may lack the resources to do such conversion, or data necessary,
or available conversion may be ambiguous.

> This will all work fine for new clients and servers but I'm having some
> problems figuring out when the welcome message can be sent.

  "Welcome message" is an optional multiline part of the successful login
message ("230 User logged in, proceed."). Nothing can prevent the issue of
the login message without breaking the protocol and compatibility with
existing software, and disabling it merely because of lack of "LANG" will
confuse users -- FTP, as opposed to HTTP, is a very old protocol, and it
has no place for a version number in the request. So existing clients that
never send any additional commands will simply be left without any
possibility to see "welcome message". Enabling this feature only of FEAT
is requested may seem to be more reasonable, but it will give FEAT a
rather ugly side effect.

  Along with "welcome messages" it's common for FTP servers to display
help, various multiline information in response to CWD, etc., so
implementation should affect either all content-dependent responses over
control socket (in the cases when translation of those particular messages
is available), or none, to be consistent.

> 
> For new clients and servers: MLST recommends the server wait until after a
> HOST command is issued to send the "welcome". What if a HOST command is not
> sent? Should the I18N draft require MLST be supported and therefore HOST
> will be supported (presumably after the LANG command and before the USER
> command). Also some discussion has recommended defining a new command for
> displaying the "welcome" message. Should a command be defined to do this and
> relieve the dependence on HOST? That latter approach seems cleaner to me.
> The requirement would be that the client issue a new "MESS" command just
> prior to authentication.

  I don't understand how proposed HOST command complies with the
requirement of backward compatibility at all. It is made as an imitation
of HTTP "Host" header, however in HTTP it became mandatory in HTTP 1.1.
Not being mandatory the possibility of HOST provides no means of reliably
identifying the virtual host if multiple hosts share the same IP address
for clients that are not aware of it, except by sending some kind of 
"welcome message" that tells the user to enter HOST manually, immediately
after his login.

  My personal opinion is that FTP should not imitate HTTP in areas where
HTTP implements its features in a way, significantly different from FTP.
If such features are necessary, HTTP should be used instead of FTP in the
first place. While it's true that FTP "competes" with HTTP, there are a
lot of situations where the use of HTTP is completely justified, and there
are absolutely no benefits that FTP provides over HTTP. In those cases
it's just more reasonable to use HTTP and leave FTP alone.

  FTP is a protocol, not a product, and increasing its use at the expense
of other protocol in the manner that will not benefit efficiency and
usability of the Internet is not what the development of Internet
standards is about. One can use FTP instead of SMTP or NNTP, too, however
currently it's clearly visible that it will be pointless to "improve" FTP
into the direction that will allow to replace those protocols because SMTP
and NNTP serve their purposes better than any "improved" FTP ever will be,
unless it will incorporate them. I propose to leave HTTP alone with its
use, incorporate into FTP the best ideas that fit (this is why long ago I
proposed to define operations to be byte-transparent by default, and to
add optional language/charset labels to MLST), and make something useful
from it. The goal of making FTP into better HTTP than HTTP is childish,
and unlikely even to benefit FTP servers and clients' developers, not to
mention users.

  There is a possibility that in the future the improvements in HTTP will
make HTTP match every piece of FTP functionality, and will make FTP
obsolete. If that will happen, nothing will be lost because functionality
will be still there. Currently this is not the case, and FTP use is
justified. I see the goal of FTP improvement not to compete with anything
or provide a possibility of unusual uses for FTP that are already served
by other protocols -- the goal is to improve FTP to allow its use in
existing situations where FTP, for good or bad reasons, is used, and will
be used for a long time, and in areas where nothing serves the purpose
well, but FTP can fit if improved. Non-ASCII filenames and directory lists
formats are among cases where FTP standards need improvements because of
the nature of data and operations, commonly performed over FTP. Virtual
hosting of web sites and file archives is the area where HTTP fits very
well, and needs no replacement -- people who want their file archive to be
on some virtual host shared with other archives, have no need in
"nice-looking" FTP URL if the price is any decrease in compatibility with
existing FTP clients. It's not HTTP where the URL became a symbol of the
identity and status of its owner.

  Anyone who wants to answer to this and argue that HTTP is bad,
inefficient, etc., should read relevant standards carefully first -- I
did, and I expect others to do the same if they want to discuss it. The
flaws of existing implementations of HTTP are well known, and I will not
argue that poor implementation of HTTP, or the enormous size of some
specifications, associated with HTTP, do not cause damage. However it's a
poor justification for extending FTP (thus bloating the standard and
creating possible incompatibilities, too) into that area.

> For old clients and new servers: How should the server recognize that it is
> talking to an old client and that it should probably issue the "Welcome"
> after the TCP connection??

  The message after TCP connection is "220 Service ready for new user." It
must be sent by the server without any expected input, and protocol that
does not implement this is by definition not FTP.

> If it waits for the LANG, HOST, MESS commands to
> first send the "welcome" message it will receive the USER command and the
> human associated with the user-PI may miss login information from the
> server.

  Only commands, sent between the initial connection and USER can affect
the login message.

-- 
Alex

----------------------------------------------------------------------
 Excellent.. now give users the option to cut your hair you hippie!
                                                  -- Anonymous Coward





From ftp-wg-owner@hethmon.com  Thu Jan 14 03:57:21 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id DAA04468
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 03:57:20 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 03:47:57 -0500
Received: from munnari.OZ.AU (munnari.oz.au [128.250.1.21] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 03:47:55 -0500
Received: from mundamutti.cs.mu.OZ.AU ([128.250.1.5])
	by munnari.OZ.AU with SMTP (5.83--+1.3.1+0.57) id IA25067;
	Thu, 14 Jan 1999 19:47:47 +1100 (from kre@munnari.OZ.AU)
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Message-Id: <25603.916303666@munnari.OZ.AU>
Date: Thu, 14 Jan 1999 03:47:55 -0500
X-OldDate:   Thu, 14 Jan 1999 19:47:46 +1100
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Robert Elz <kre@munnari.OZ.AU>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST nits

I am just cleaning up some oddments in my MLST implementation
- done mostly so I can include some real examples in the draft,
rather than to be useful for anyone - it is littered with code
like
	static char pbuf[1040];
which should never see any real use anywhere, but it works for my
private testing (if I wanted to break it, I have other ways than
concocting long strings, etc).

Anyway, while doing it, I encountered a few questions.  Most of these
don't necessarily involve anything that would change the mlst draft
in any way, though in a few, if we get soem kind of clear agreement
that additions to, or changes to, the text are needed, I will do that.
(A new version of the draft with more examples will appear sometime
soon anyway).

The first, and perhaps the most likely to perhaps result in a change
relates to the following text in the draft...

   The "l" permission applies to the directory file types, and indicates
   that the listing commands, LIST, NLST, and MLSD may be applied to the
   directory in question, and that MLST, LIST, NLST, and STAT may be
   applied to objects in the directory.

The relevant part is the "and that ..." clause at the end.   Perhaps this
is obvious, but it means that a directory which can be listed, but for
which the contents can't be referenced at all (on unix, a directory with
444 mode for example) can't be indicated as being listable, even though it
is.   Now such things are quite rare, almost useless, and perhaps deserve
to not be marked as being listable, in which case everything can stay just
the same as it is (including my code, which currently tests for both read
and exec perm bits before it sets the 'l' indicator).

Or, perhaps we might split those two features into separate indicators,
keep 'l' for what it is, and make a new one for "information can be obtained
about objects in the directory".   On unix the 'e' flag has the same
meaning as this new flag would have, but I would not like to overload it
in general (that is, for a unix server, 'e' and 'new' would both be set
when either was, just as 'c' 'p' and 'm' are either all set, or all clear.
On some other underlying filesys, 'l' and 'new' might always both be set
or clear together - all of that is OK).

Opinions?

Next, I hope it is clear in the draft already that implementations are not
required to report any facts which make no sense to report, and that that
is a file by file decision that can be made, not just a "my implementation
has no way to make the unique fact, so I never report it" type of decision
(though that one is acceptable as well).   If anyone wasn't clear about that
then let us know, so the draft can be made clearer - if you can point to a
place in the draft which suggests otherwise, or where more clarity could be
added, that would help.

The query relates to reporting the size of a directory object.   I was just
reporting the st_size value that unix supplies (number of bytes occupied by
the directory on disc) - but this really makes no sense at all, it is a number
that relates to almost nothing (except approximately the sum of the lengths of
the filenames the directory once contained, typically rounded up).   Now, my
implementation omits the size fact for directories (and consequently unless
I alter it, so will the examples that go in the draft).

The question is whether anything should be said in the draft about how the
size fact relates to directories?   It could be recommended to be omitted
(said not to apply to objects of type dir pdir or cdir), or it could be
defined in some other way for directories (perhaps a count of the number of
entries in the directory, or something).

Opinions?  Comments?

I also omit the size for objects that happen to be unix block or character 
special files - not that I couldn't (sometimes) find a reasonable size to
return there (discs have sizes anyway), but I'm too lazy, permitted to
omit the size, and so I will...   A really good implementation probably would
not, but mine is not that.

These however raise another question entirely.

We already have in the draft ...

   Files types that are specific to a specific operating system, or file
   system, can be encoded using the "OS." type names.  The format is:

        os-type   = "OS." os-name "=" localtype
        os-name   = <an IANA registered operating system name>
        localtype = 1*RCHAR

which is fine - I can, and do, return OS.UNIX=...

(UNIX is one of the IANA registered OS names, and the most appropriate for
this purpose, as block and char special files etc are common to all of the
unix variants I have ever heard of).

The question now is whether we need any form of commonality or registration
for the "localtype".   We can just ignore this, and allow the market to
reach some common usages as required, or we could request the IANA to make
a registry of OS dependent types (I don't think we want to attempt to
list everything here - that would be insane).

What I do currently is ...

Type=OS.unix=chr-13/29;Modify=19990104143839;Perm=;Unique=keVO1+5G4 device

(and yes, that is an output line from my test environment, "device" is the
file name of the char dev I created for the test it has no permissions
because I didn't care what major/minor numbers I gave it, and wanted to
make sure I didn't accidentally damage something by reading or writing
whatever it happens to be ... ie: I did a "chmod 0 device" or something).

The "chr-13/29" is supposed to indicate character device, major 13,
minor 29.   The same is done for block devices (using "blk") - I'm not
doing the "OS.Unix=BLOCK" thig which is in the draft purely as a means
to illustrate that the "localtype" is case independent.

So, do we want to request of the IANA that they register these OS
dependent file types or not?   If we do, I would expect that all that
should be needed for a registration would be message from someone
giving the OS name, the localtype they're registering, and a definition
for what it means.   That is, the idea would be just to make the information
available, not for anyone particular to debate it (the IANA would
verify that there weren't dozens of registrations of the same thing).

The same should be asked of the OS dependent facts.   My FEAT output
says (currently anyway) ...

 MLST Type*;Size*;Modify*;Perm*;Unique*;UNIX.mode;UNIX.chgd

The last two (which are off by default) are OS specific.  When I turn them
on (and turn off most others to keep the line here shorter), for the
same device above, I get ...

 Type=OS.unix=chr-13/29;UNIX.mode=020400;UNIX.chgd=19990104145937 device

(that's actually from MLST, hence the leading space).

UNIX.mode is the st_mode value (as is - so you can see I actually made the
device 0400 mode, it will be owned by root), UNIX.chgd is the "st_ctime"
value - which indicates when (almost) anything associated with the file was
last altered (which can be useful for mirroring type applications, perhaps).
Unix.mode has an octal integer as its value (standard unix representation),
and UNIX.chgd has a time-val as its value.

In the FEAT output above you'll see that I don't support the lang, charset,
or media-type facts (though they are recognised by the command parser as
valid fact names, so if you try "OPTS MLST charset" it won't be an error,
whereas "OPTS MLST foobar" would be - the "charset" fact request will just
be ignored).   This is all as the draft says should be done I believe.

The OS specific facts come from the text in the draft ...

   Further operating system specific keywords could be specified by
   using the IANA operating system name as a prefix (examples only):

        OS/2.ea   -- OS/2 extended attributes
        MACOS.rf  -- MacIntosh resource forks
        UNIX.mode -- Unix file modes (permissions)

The question is whether this is enough of a specification of OS specific
facts, and whether we should ask the IANA to maintain a registry of these
kinds of facts?

I believe that there are one or two other implementations being done.
Do the people doing those (if they're on this list) have any other issues
or comments to raise?

If so, now would be an excellent time to make comments.   I'll be doing an
update of the draft, including MLST examples, soon - if there aren't any
more issues to resolve, then that might very well be the last of the mlst
draft series.

kre




From ftp-wg-owner@hethmon.com  Thu Jan 14 08:10:56 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id IAA05725
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 08:10:55 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 08:07:04 -0500
Received: from deliverator.io.com (deliverator.io.com [199.170.88.17] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 08:07:02 -0500
Received: from mx2.io.com (IDENT:root@mx2.io.com [199.170.88.18])
	by deliverator.io.com (8.9.1a/8.9.1a) with ESMTP id HAA02582
	for <ftp-wg@hethmon.com>; Thu, 14 Jan 1999 07:05:25 -0600
Received: from magnus (as1-dialup-93.io.com [206.224.82.93])
	by mx2.io.com (8.9.1a/8.9.1a) with SMTP id HAA09404;
	Thu, 14 Jan 1999 07:06:56 -0600
Message-Id: <3.0.5.32.19990114064048.012fa0b0@mail.io.com>
X-Sender: alun@mail.io.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 3.0.5 (32)
In-Reply-To: <25603.916303666@munnari.OZ.AU>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Thu, 14 Jan 1999 08:07:02 -0500
X-OldDate:   Thu, 14 Jan 1999 06:40:48 -0600
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Alun Jones <alun@texis.com>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST nits

At 03:47 AM 1/14/99 -0500, Robert Elz wrote:
>The first, and perhaps the most likely to perhaps result in a change
>relates to the following text in the draft...
>
>   The "l" permission applies to the directory file types, and indicates
>   that the listing commands, LIST, NLST, and MLSD may be applied to the
>   directory in question, and that MLST, LIST, NLST, and STAT may be
>   applied to objects in the directory.
>
>The relevant part is the "and that ..." clause at the end.   Perhaps this
>is obvious, but it means that a directory which can be listed, but for
>which the contents can't be referenced at all (on unix, a directory with
>444 mode for example) can't be indicated as being listable, even though it
>is.   Now such things are quite rare, almost useless, and perhaps deserve
>to not be marked as being listable, in which case everything can stay just
>the same as it is (including my code, which currently tests for both read
>and exec perm bits before it sets the 'l' indicator).

Correct me if I'm wrong, but such a directory isn't listable by MLSD,
either, right?  Hence, it's not markable with an 'l' in the first place.
As a side-issue, of course, some file systems (FAT comes to mind) have the
concept of a 'hidden' file, which should not be listed, even if the
directory is listable.  The 'l' shouldn't suggest that all these hidden
files are listable.

>Next, I hope it is clear in the draft already that implementations are not
>required to report any facts which make no sense to report, and that that
>is a file by file decision that can be made, not just a "my implementation
>has no way to make the unique fact, so I never report it" type of decision
>(though that one is acceptable as well).   If anyone wasn't clear about that
>then let us know, so the draft can be made clearer - if you can point to a
>place in the draft which suggests otherwise, or where more clarity could be
>added, that would help.

I'm sure that, however clear we make it, someone will write a client that
expects exactly the same range (and ordering) of facts to be presented on
each line of an MLSD/MLST output.

>The question is whether anything should be said in the draft about how the
>size fact relates to directories?   It could be recommended to be omitted
>(said not to apply to objects of type dir pdir or cdir), or it could be
>defined in some other way for directories (perhaps a count of the number of
>entries in the directory, or something).
>
>Opinions?  Comments?

In the past, we've seen that some items that are undefined (such as listing
formats) will always be expected to match a de-facto standard.  In this
case, we have the advantage that not supplying the fact is a valid response
- since it is just as valid to suggest that the size reported be a total of
the sizes of each item in the directory (which could give away the presence
of hidden files!), as it is to report the number of files within the
directory (again - only non-hidden files?), maybe we should state that the
size fact does not apply to directories, and clients should not expect it
to appear, or carry meaningful information if it does.

>So, do we want to request of the IANA that they register these OS
>dependent file types or not?   If we do, I would expect that all that
>should be needed for a registration would be message from someone
>giving the OS name, the localtype they're registering, and a definition
>for what it means.   That is, the idea would be just to make the information
>available, not for anyone particular to debate it (the IANA would
>verify that there weren't dozens of registrations of the same thing).

That sounds like it would be useful, although it should be emphasised that
the creation of such facts should generally be in some manner related to
the OS, and not for the creation of pseudo-facts to reflect a 'virtual file
system' or some other artifact that the FTP server builds to supplement or
supplant a feature of the OS.  Is there an OS type that could be used for
such non-OS features?  I know that often a feature (such as file ownership
on FAT volumes) will be mimicked to provide similarity to other systems,
despite that such a feature is not provided by the OS.

>The question is whether this is enough of a specification of OS specific
>facts, and whether we should ask the IANA to maintain a registry of these
>kinds of facts?

If they are to be usefully registered at all, then the format of those
facts must be registered as well.

I'll be taking a look back at the HOST section in the MLST draft shortly
and let you know if I have any further comments on it.

Alun.
~~~~

--
Texas Imperial Software | Try WFTPD, the Windows FTP Server. Find it
1602 Harvest Moon Place | at web site http://www.wftpd.com or email
Cedar Park TX 78613     | us at alun@texis.com.  VISA / MC accepted.
Fax +1 (512) 378 3246   | NT based ISPs, be sure to read details of
Phone +1 (512) 378 3246 | WFTPD Pro, NT service version - $100.
*WFTPD and WFTPD Pro now available as native Alpha versions for NT*



From ftp-wg-owner@hethmon.com  Thu Jan 14 08:11:12 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id IAA05740
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 08:11:11 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 08:07:12 -0500
Received: from deliverator.io.com (deliverator.io.com [199.170.88.17] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 08:07:10 -0500
Received: from mx2.io.com (IDENT:root@mx2.io.com [199.170.88.18])
	by deliverator.io.com (8.9.1a/8.9.1a) with ESMTP id HAA02591
	for <ftp-wg@hethmon.com>; Thu, 14 Jan 1999 07:05:32 -0600
Received: from magnus (as1-dialup-93.io.com [206.224.82.93])
	by mx2.io.com (8.9.1a/8.9.1a) with SMTP id HAA09409
	for <ftp-wg@hethmon.com>; Thu, 14 Jan 1999 07:07:04 -0600
Message-Id: <3.0.5.32.19990114070136.012fbbd0@mail.io.com>
X-Sender: alun@mail.io.com
X-Mailer: QUALCOMM Windows Eudora Pro Version 3.0.5 (32)
In-Reply-To: <Pine.LNX.4.02A.9901121307080.30532-100000@phobos.illtel.de
 nver.co.us>
References: <43B821CCD144D211AB0500204804EE4A2BB21C@rbmail102.chamb.disa.mil>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Thu, 14 Jan 1999 08:07:10 -0500
X-OldDate:   Thu, 14 Jan 1999 07:01:36 -0600
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Alun Jones <alun@texis.com>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   Lang command

At 05:52 PM 1/12/99 -0500, you wrote:
>  I don't understand how proposed HOST command complies with the
>requirement of backward compatibility at all. It is made as an imitation
>of HTTP "Host" header, however in HTTP it became mandatory in HTTP 1.1.
>Not being mandatory the possibility of HOST provides no means of reliably
>identifying the virtual host if multiple hosts share the same IP address
>for clients that are not aware of it, except by sending some kind of 
>"welcome message" that tells the user to enter HOST manually, immediately
>after his login.

That is correct.  The intent is that sites in future will be of the
following kinds:

1. RFC-959 compliant: no HOST command is accepted, and either FEAT is
unrecognised, or does not return HOST as an option.  The client is not
expected (or allowed) to send a HOST command.
2. HOST command accepted, FEAT reports that HOST is allowed, yet there is
some 'default' environment/host that is entered by those clients with no
HOST command implemented.  Here, the greeting message could make it clear
that the "QUOTE HOST" command can be used to enter other hosts.
3. HOST command required - FEAT reports that HOST is allowed, no other
action is allowed until the HOST command is entered.

The FTP server will still send a 220 message in greeting, and a 230 message
once the user and password have been verified on that host.  No difference
there.  I've suggested, I believe, that the response to HOST be 220, so
that a HOST command could be implemented by a wrapper that passes the
client on to a standard server.

>  My personal opinion is that FTP should not imitate HTTP in areas where
>HTTP implements its features in a way, significantly different from FTP.
>If such features are necessary, HTTP should be used instead of FTP in the
>first place. While it's true that FTP "competes" with HTTP, there are a
>lot of situations where the use of HTTP is completely justified, and there
>are absolutely no benefits that FTP provides over HTTP. In those cases
>it's just more reasonable to use HTTP and leave FTP alone.

The HOST command was suggested from the simple standpoint that many people
have been asking for it.  Yes, it derives in some sense from the perception
among users that if HTTP can distinguish between hosts on the same IP
address, surely FTP should also.  No, it isn't an attempt to compete with
or supplant FTP - it is merely an attempt to allow co-location of servers
without further depleting the IP address space by requiring a different IP
address for each FTP server.  To some extent, it is necessitated by the
fact that multiple HTTP servers can run on the same IP address and be
distinguished from one another, and that FTP servers may be dropped if they
cannot do so.  To lose FTP's numerous advantages over HTTP for this reason
would seem petty.

>from it. The goal of making FTP into better HTTP than HTTP is childish,
>and unlikely even to benefit FTP servers and clients' developers, not to
>mention users.

This is not even remotely what this is about.  You've got the wrong
impression entirely.  We are not attempting to supplant HTTP - although
your same accusation might be thrown at HTTP (after all, FTP had a
persistent connection for multiple transfers _long_ before HTTP did it, and
for HTTP to do it just because FTP has it "is childish", etc, right?)

>  There is a possibility that in the future the improvements in HTTP will
>make HTTP match every piece of FTP functionality, and will make FTP
>obsolete. If that will happen, nothing will be lost because functionality
>will be still there.

You sound like you're saying that HTTP should continue to improve to
subsume FTP, and that FTP should remain static?  I don't accept that.
There is still a place for FTP, and there is still room for improvement to
FTP.  A protocol that refuses to improve to match bona fide requirements is
doomed.

>  Virtual
>hosting of web sites and file archives is the area where HTTP fits very
>well, and needs no replacement -- people who want their file archive to be
>on some virtual host shared with other archives, have no need in
>"nice-looking" FTP URL if the price is any decrease in compatibility with
>existing FTP clients. It's not HTTP where the URL became a symbol of the
>identity and status of its owner.

Excuse me?  You're now saying that a file archive is the perfect use of
HTTP, and somewhere that FTP shouldn't stick its nose in, for the sole
reason that the archive is 'virtual'?  This same effect can be achieved
right now, through FTP as it exists, with the major downside that it uses
one more IP address, and requires each server that is co-located on a
machine to have its own IP address.  The addition of the HOST command
serves solely to remove that requirement of multiple IP addresses per
machine (which is often difficult to achieve in some OSes).

>  The message after TCP connection is "220 Service ready for new user." It
>must be sent by the server without any expected input, and protocol that
>does not implement this is by definition not FTP.

This remains true with the HOST command, as for the LANG command.  A 220
greeting message is _always_ to be issued to the client on connection.

>> If it waits for the LANG, HOST, MESS commands to
>> first send the "welcome" message it will receive the USER command and the
>> human associated with the user-PI may miss login information from the
>> server.
>
>  Only commands, sent between the initial connection and USER can affect
>the login message.

My original idea is for the HOST command to send a fresh greeting, based on
the server's newly selected identity.  If HOST and LANG are both supported,
and both generate a greeting message, this could be an issue that needs
fixing - either allowing a language argument on the HOST command, for
instance, or by stating that one should precede the other.

Alun.
~~~~

--
Texas Imperial Software | Try WFTPD, the Windows FTP Server. Find it
1602 Harvest Moon Place | at web site http://www.wftpd.com or email
Cedar Park TX 78613     | us at alun@texis.com.  VISA / MC accepted.
Fax +1 (512) 378 3246   | NT based ISPs, be sure to read details of
Phone +1 (512) 378 3246 | WFTPD Pro, NT service version - $100.
*WFTPD and WFTPD Pro now available as native Alpha versions for NT*



From ftp-wg-owner@hethmon.com  Thu Jan 14 09:25:10 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id JAA06134
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 09:25:10 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 09:21:26 -0500
Received: from ACFcluster.NYU.EDU (axp2.acf.nyu.edu [128.122.250.38] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 09:21:24 -0500
Received: from ACFcluster.NYU.EDU by ACFcluster.NYU.EDU (PMDF V5.2-29 #33915)
 id <01J6IXHRAM3K8WVYO2@ACFcluster.NYU.EDU> for ftp-wg@hethmon.com; Thu,
 14 Jan 1999 09:21:17 EDT
In-reply-to: "Your message dated Thu, 14 Jan 1999 08:07:02 -0500"
 <3.0.5.32.19990114064048.012fa0b0@mail.io.com>
Message-id: <01J6IXWDSSA28WVYO2@ACFcluster.NYU.EDU>
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii
References: <25603.916303666@munnari.OZ.AU>
Date: Thu, 14 Jan 1999 09:21:25 -0500
X-OldDate:   Thu, 14 Jan 1999 09:20:36 -0400 (EDT)
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Stephen Tihor <TIHOR@ACFcluster.NYU.EDU>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST nits

> >Next, I hope it is clear in the draft already that implementations are not
> >required to report any facts which make no sense to report, and that that
> >is a file by file decision that can be made, not just a "my implementation
> >has no way to make the unique fact, so I never report it" type of decision
> >(though that one is acceptable as well).   If anyone wasn't clear about that
> >then let us know, so the draft can be made clearer - if you can point to a
> >place in the draft which suggests otherwise, or where more clarity could be
> >added, that would help.

> I'm sure that, however clear we make it, someone will write a client that
> expects exactly the same range (and ordering) of facts to be presented on
> each line of an MLSD/MLST output.

Probably. But its one reason to include examples where the range and ordering
of the facts varies arbitrarilly from line to line :)



From ftp-wg-owner@hethmon.com  Thu Jan 14 10:01:05 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id KAA06585
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 10:01:04 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 09:57:13 -0500
Received: from munnari.OZ.AU (munnari.oz.au [128.250.1.21] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 09:57:11 -0500
Received: from mundamutti.cs.mu.OZ.AU ([128.250.1.5])
	by munnari.OZ.AU with SMTP (5.83--+1.3.1+0.57) id OA29814;
	Fri, 15 Jan 1999 01:57:03 +1100 (from kre@munnari.OZ.AU)
In-Reply-To: Your message of "Thu, 14 Jan 1999 08:07:02 CDT."
             <3.0.5.32.19990114064048.012fa0b0@mail.io.com> 
Message-Id: <170.916325823@munnari.OZ.AU>
Date: Thu, 14 Jan 1999 09:57:11 -0500
X-OldDate:   Fri, 15 Jan 1999 01:57:03 +1100
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Robert Elz <kre@munnari.OZ.AU>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG: MLST nits 

    Date:        Thu, 14 Jan 1999 08:07:02 -0500
    From:        Alun Jones <alun@texis.com>
    Message-ID:  <3.0.5.32.19990114064048.012fa0b0@mail.io.com>

  | Correct me if I'm wrong, but such a directory isn't listable by MLSD,
  | either, right?

My server will list it correctly if no facts are requested (none
at all).   In that case it doesn't need to stat() the files in the
directory, and so doesn't.   Then read permission is enough.

It acts a little oddly if any facts are requested, acting almost
as if the directory is empty...

Thus ...

---> PORT 127,0,0,1,4,188
200 PORT command successful.
---> MLSD .
150 BINARY connection open for MLSD .
Type=pdir;Modify=19990104143638;Perm=e;Unique=keVO1+d?3 ..
Type=file;Size=0;Modify=19990104143932;Perm=awr;Unique=keVO1+8G4 writeable
Type=dir;Modify=19990114065955;Perm=cpmel;Unique=keVO1+7G4 promiscuous
Type=dir;Modify=19990114142538;Perm=;Unique=keVO1+1t2 no-exec
226 MLSD completed

Note no-exec which has no perms (no 'l' as, as currently implemented,m
it requires that the directory have the unix 'x' bit).   (There were a
few more output lines there which aren't relevant, snd so I deleted).

---> PORT 127,0,0,1,4,193
200 PORT command successful.
---> MLSD no-exec
150 BINARY connection open for MLSD no-exec
Type=cdir;Modify=19990114142538;Perm=;Unique=keVO1+1t2 no-exec
226 MLSD completed

That's with the default fact set (that was shown in the FEAT output
of my last message).

If then ...

---> opts mlst
200 Done
ftp> mlsd no-exec
---> PORT 127,0,0,1,4,194
200 PORT command successful.
---> MLSD no-exec
150 BINARY connection open for MLSD no-exec
 a
 b
 c
 d
226 MLSD completed

This is kind of similar to what a LIST command does...

---> PORT 127,0,0,1,4,211
200 PORT command successful.
---> LIST no-exec
150 Opening ASCII mode data connection for /bin/ls.
no-exec/a: No permission
no-exec/b: No permission
no-exec/c: No permission
no-exec/d: No permission
total 0
226 Transfer complete.

Pretty weird ...  but you still get to discover what files are
in the directory.   If I use a different client that I can (pretty
stupidly) convince to use the NLST command for other than an MGET
or similar, what I see is ..

200 PORT command successful.
---> NLST no-exec
150 Opening ASCII mode data connection for file list.
no-exec/a
no-exec/b
no-exec/c
no-exec/d
226 Transfer complete.

Then if I go back to my newer client ...

---> opts mlst type;modify;perm
200 Done

---> PORT 127,0,0,1,4,225
200 PORT command successful.
---> MLSD promiscuous
150 BINARY connection open for MLSD promiscuous
Type=cdir;Modify=19990114065955;Perm=cpmel promiscuous
Type=pdir;Modify=19990114142445;Perm=el ..
226 MLSD completed


That's what a listing of an empty directory would normally look like.
With no facts requested ...

---> opts mlst
200 Done
---> PORT 127,0,0,1,4,219
200 PORT command successful.
---> MLSD promiscuous
150 BINARY connection open for MLSD promiscuous
226 MLSD completed

Note no Type=cdir or type=pdir output there, as without the type
fact, there's no way to recognise them for what they are (there
might be an argument for keeping the type=pdir in this case, as it
is in fact an entry in the directory, the type=cdir cannot be
left in the output - perhaps I should make that explicit in the
doc?

(Aside: the reason the relative paths in the Type=cdir, rather than
fully qualified is because in this particular location in my ftp
tree, the server cannot determine fully qualified path names, it
lacks permission to do so.   Elsewhere...

---> PORT 127,0,0,1,4,233
200 PORT command successful.
---> MLSD .
150 BINARY connection open for MLSD .
Type=cdir;Modify=19990112030453;Perm=el /iana
Type=pdir;Modify=19990112030508;Perm=el ..
Type=dir;Modify=19990114073438;Perm=el assignments
226 MLSD completed

---> MLST assignments
250- Listing assignments
 Type=dir;Modify=19990114073438;Perm=el /iana/assignments
250 End

That comment is just for anyone who wondered).

  | Hence, it's not markable with an 'l' in the first place.

This may very well be the best solution though, certainly the
listing, even if available, isn't terribly useful!

  | As a side-issue, of course, some file systems (FAT comes to mind) have the
  | concept of a 'hidden' file, which should not be listed, even if the
  | directory is listable.  The 'l' shouldn't suggest that all these hidden
  | files are listable.

Does it now?    Can you suggest clarifying text?

  | I'm sure that, however clear we make it, someone will write a client that
  | expects exactly the same range (and ordering) of facts to be presented on
  | each line of an MLSD/MLST output.

I will try to avoid some of that by what is in the examples.  Unfortunately
my client never varies the order in which it presents the facts (doing
that would require some pretty weird coding I think), so I will have to
fake the responses a little to get that example.   But I think that is
worth while.   Thanks for the reminder on that one.

  | maybe we should state that the
  | size fact does not apply to directories,

That certainly suits me.   Avoids a bunch of work to implement anything
else.

  | That sounds like it would be useful, although it should be emphasised that
  | the creation of such facts should generally be in some manner related to
  | the OS,

OK.

  | and not for the creation of pseudo-facts to reflect a 'virtual file
  | system' or some other artifact that the FTP server builds to supplement or
  | supplant a feature of the OS.

I'm not sure I can imagine anyoen adding features to their virtual
filestore just so they can invent a new OS dependent fact to report it.
That one seems a bit far fetched.   I can certainly see implementors
inventing data to fit the standard facts, because it seems important
to support them all (I could have pretended that all my filenames are
Lang=English;Charset=utf-8 if I wanted to (or Charset=ascii).   That
happens to be true, but my server doesn't really have any means of
knowing it, the data is not recorded anywhere - not implementing the
facts is a better choice.   I will try to make that clear in the examples.

  | Is there an OS type that could be used for| such non-OS features?

Not that I recall from my scan of the OS type listing.   But someone can
always ask the IANA to add one...

  | I know that often a feature (such as file ownership
  | on FAT volumes) will be mimicked to provide similarity to other systems,
  | despite that such a feature is not provided by the OS.

Fortunately, we have no concept of ownership in the FTP NVFS (with or
without TVFS), so there's no problem of that one with MLST at least.
(I don't think it makes any sense for implementors to even add OS
specific facts to report that data, as over a FTP connection there
can't really be any expectation that the identity information revealed
means anything at all).

  | If they are to be usefully registered at all, then the format of those
  | facts must be registered as well.

Yes, of course.

  | I'll be taking a look back at the HOST section in the MLST draft shortly
  | and let you know if I have any further comments on it.

Thanks, for this, and for that.

kre



From ftp-wg-owner@hethmon.com  Thu Jan 14 10:18:12 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id KAA07073
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 10:18:11 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 10:14:15 -0500
Received: from munnari.OZ.AU (munnari.oz.au [128.250.1.21] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 10:14:13 -0500
Received: from mundamutti.cs.mu.OZ.AU ([128.250.1.5])
	by munnari.OZ.AU with SMTP (5.83--+1.3.1+0.57) id PA29849;
	Fri, 15 Jan 1999 02:14:07 +1100 (from kre@munnari.OZ.AU)
In-Reply-To: Your message of "Thu, 14 Jan 1999 09:57:11 CDT."
             <170.916325823@munnari.OZ.AU> 
Message-Id: <310.916326846@munnari.OZ.AU>
Date: Thu, 14 Jan 1999 10:14:13 -0500
X-OldDate:   Fri, 15 Jan 1999 02:14:06 +1100
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Robert Elz <kre@munnari.OZ.AU>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG: MLST nits 

    Date:        Thu, 14 Jan 1999 09:57:11 -0500
    From:        Robert Elz <kre@munnari.OZ.AU>
    Message-ID:  <170.916325823@munnari.OZ.AU>

Apologies for replying to my own mail, but ...

  | ---> PORT 127,0,0,1,4,193
  | 200 PORT command successful.
  | ---> MLSD no-exec
  | 150 BINARY connection open for MLSD no-exec
  | Type=cdir;Modify=19990114142538;Perm=;Unique=keVO1+1t2 no-exec
  | 226 MLSD completed

I have just realised that is crazy, what it should do, and now actually
does, is ...

---> PORT 127,0,0,1,5,84
200 PORT command successful.
---> MLSD no-exec
150 BINARY connection open for MLSD no-exec
Type=cdir;Modify=19990114142538;Perm=;Unique=keVO1+1t2 no-exec
Type=pdir ..
 a
 b
 c
 d
226 MLSD completed


That is, it lists all the facts it can list for each file, omiting
any that it can't find.

So, should the 'l' flag still be off for my no-exec directory (as you
can see above that it is), or should it be on, with some other flag
indicating that information about entries in the directory will be
available?   Obviously it doesn't matter mush in this particular case,
it is obvious that no info is available, but if one does ..

---> mlst no-exec
250- Listing no-exec
 Type=dir;Modify=19990114142538;Perm=;Unique=keVO1+1t2 no-exec
250 End

One might like to know that listing the directory won't be
terribly enlightening, but is possible?

Similarly, from a listing of the parent directory.

kre



From ftp-wg-owner@hethmon.com  Thu Jan 14 13:27:17 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id NAA13447
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 13:27:16 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 13:23:15 -0500
Received: from usilws34.ca-nethaven.com (usilws34.ca-nethaven.com [206.79.17.234] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 13:23:13 -0500
Received: from mike_gleason ([208.200.66.72]) by usilws34.ca-nethaven.com
          (Netscape Messaging Server 3.01)  with SMTP id AAA20541
          for <ftp-wg@hethmon.com>; Thu, 14 Jan 1999 13:04:25 -0500
Message-Id: <4.1.19990114105020.00a46be0@206.79.17.234>
X-Sender: mgleason@206.79.17.234
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
In-Reply-To: <25603.916303666@munnari.OZ.AU>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Thu, 14 Jan 1999 13:23:13 -0500
X-OldDate:   Thu, 14 Jan 1999 12:18:07 -0600
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: "Mike Gleason" <mgleason@ncftp.com>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST nits

At 03:47 AM 1/14/99 -0500, you wrote:

>Now, my implementation omits the size fact for directories (and
consequently unless
>I alter it, so will the examples that go in the draft).

This is what NcFTPd Server (UNIX) does and it seems to work well in
practice.  NcFTP Client simply leaves that field blank when it prints
directory listings obtained via MLSD.  Although admittedly, those two use
an implementation that is several months older than the latest spec.


>The question now is whether we need any form of commonality or registration
>for the "localtype". [...] The same should be asked of the OS dependent
facts. 
>My FEAT output says (currently anyway) ...
>
> MLST Type*;Size*;Modify*;Perm*;Unique*;UNIX.mode;UNIX.chgd

FWIW, NcFTPd implements UNIX.mode, UNIX.owner, and UNIX.group.  I do think
the working group needs to help standardize these OS-specific items in some
form.  For example, NcFTPd's UNIX.mode only provides owner/group/other
(i.e. st_mode & 0777).  It also prints the mode as an octal number, using a
leading zero.  However, I have seen another implementation that simply
prints the whole st_mode in decimal.  A client wishing to use this fact
can't do this effectively if every implementation is different.

The spec should probably not directly define OS-specific facts, but if
OS-specific standards have evolved it should point to those documents when
possible.  This would cover the case where someone reads the spec but does
not participate directly in this group.

Mike Gleason





From ftp-wg-owner@hethmon.com  Thu Jan 14 13:47:16 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id NAA13835
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 13:47:15 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 13:43:21 -0500
Received: from usilws34.ca-nethaven.com (usilws34.ca-nethaven.com [206.79.17.234] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 13:43:19 -0500
Received: from mike_gleason ([208.200.66.72]) by usilws34.ca-nethaven.com
          (Netscape Messaging Server 3.01)  with SMTP id AAA20665
          for <ftp-wg@hethmon.com>; Thu, 14 Jan 1999 13:24:27 -0500
Message-Id: <4.1.19990114122308.00a72b40@206.79.17.234>
X-Sender: mgleason@206.79.17.234
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
In-Reply-To: <25603.916303666@munnari.OZ.AU>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Thu, 14 Jan 1999 13:43:19 -0500
X-OldDate:   Thu, 14 Jan 1999 12:40:42 -0600
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: "Mike Gleason" <mgleason@ncftp.com>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST nits

At 03:47 AM 1/14/99 -0500, you wrote:

>The question now is whether we need any form of commonality or registration
>for the "localtype".  [...]
>
>Type=OS.unix=chr-13/29;Modify=19990104143839;Perm=;Unique=keVO1+5G4 device
>
>(and yes, that is an output line from my test environment, "device" is the
>file name of the char dev I created for the test

I may have dozed off during discussion of this, but do we really need OS
specific types at all?  FTP clients and their users probably don't care
about device "files", etc.  IMHO, for FTP there are only two useful items
to the client, a data container (directory/folder) or a data object (file).
 If a user wants to see device files, socket files, named pipes, etc., let
them login by telnet.

Even an element in the filesystem that could be useful to FTP like a named
pipe could still be represented as a "file" if you can read from it or
write to it.  (However, in that case it could have indeterminate size, so
no size fact would be reported.)  Otherwise, you can't do anything to the
object otherwise, so from the user's perspective, it might as well not
exist (and therefore not be listed).

I apologize if my sleep deprivation is causing me to be short-sighted and
miss something blatantly obvious...




From ftp-wg-owner@hethmon.com  Thu Jan 14 18:37:33 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id SAA19127
	for <ftpext-archive@lists.ietf.org>; Thu, 14 Jan 1999 18:37:32 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 18:31:58 -0500
Received: from munnari.OZ.AU (munnari.oz.au [128.250.1.21] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Thu, 14 Jan 1999 18:31:56 -0500
Received: from mundamutti.cs.mu.OZ.AU ([128.250.1.5])
	by munnari.OZ.AU with SMTP (5.83--+1.3.1+0.57) id XA04397;
	Fri, 15 Jan 1999 10:28:33 +1100 (from kre@munnari.OZ.AU)
In-Reply-To: Your message of "Thu, 14 Jan 1999 13:43:19 CDT."
             <4.1.19990114122308.00a72b40@206.79.17.234> 
Message-Id: <3336.916356507@munnari.OZ.AU>
Date: Thu, 14 Jan 1999 18:31:56 -0500
X-OldDate:   Fri, 15 Jan 1999 10:28:27 +1100
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Robert Elz <kre@munnari.OZ.AU>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG: MLST nits 

    Date:        Thu, 14 Jan 1999 13:43:19 -0500
    From:        "Mike Gleason" <mgleason@ncftp.com>
    Message-ID:  <4.1.19990114122308.00a72b40@206.79.17.234>

  | I may have dozed off during discussion of this, but do we really need OS
  | specific types at all?

I suspect they're useful for mirroring, and similar.   FTP is more
than anon FTP (or ought to be).

  | Otherwise, you can't do anything to the
  | object otherwise, so from the user's perspective, it might as well not
  | exist (and therefore not be listed).

You may be able to recreate it, if what you're trying to so is make
a direct image of a filesys (a mirror).  True, named pipes and sockets
of various types aren't likely to be useful that way, but other things
can be (including devices).   I see less (but not zero) application for
the owner/group info than for being able to list the entire filesys
some way or other.

It seems that (so far anyway) there is agreement that we ought ask the
IANA to maintain a registry of OS specific facts and their definitions,
and from that, and if we do not agree to remove OS specific file types,
I'd assume (so far) agreement to ask the IANA to create a registry for
those as well.

I will write some text for the IANA considerations section, probably
over the weekend, and send it to the list for comment.

kre



From ftp-wg-owner@hethmon.com  Fri Jan 15 00:14:29 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id AAA25549
	for <ftpext-archive@lists.ietf.org>; Fri, 15 Jan 1999 00:14:23 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Fri, 15 Jan 1999 00:06:47 -0500
Received: from ACFcluster.NYU.EDU (axp2.acf.nyu.edu [128.122.250.38] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Fri, 15 Jan 1999 00:06:37 -0500
Received: from ACFcluster.NYU.EDU by ACFcluster.NYU.EDU (PMDF V5.2-30 #33915)
 id <01J6JRF7M4RK8WVYNG@ACFcluster.NYU.EDU> for ftp-wg@hethmon.com; Fri,
 15 Jan 1999 00:06:16 EDT
In-reply-to: "Your message dated Thu, 14 Jan 1999 13:43:19 -0500"
 <4.1.19990114122308.00a72b40@206.79.17.234>
Message-id: <01J6JSTL6GMA8WVYNG@ACFcluster.NYU.EDU>
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii
References: <25603.916303666@munnari.OZ.AU>
Date: Fri, 15 Jan 1999 00:06:38 -0500
X-OldDate:   Fri, 15 Jan 1999 00:00:21 -0400 (EDT)
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Stephen Tihor <TIHOR@ACFcluster.NYU.EDU>
To: Mike Gleason <mgleason@ncftp.com>
Subject: Ftp-WG:   MLST nits

Some data objects are quite complex.  For file system mirroring lots of
additional information needs to be moved around.  For a elegant transmission of 
complex objects (multipart files with complex typing information outside the 
pure body) it is also useful since one can use a much more natural
decomposition between header and body.   For FTP based network file systems it
is very useful indeed. 

There are other much less elegant ways to hack around these issues but they are
just hacks.   Even in the simpel UNIX space clean handling of hard and soft
links benefits from OS specific facts. 

Registering their semantics seems a good idea.  Pre-registering a setof POSIX
facts is probably politically correct and a useful suggestion for how to handle
most common UNIXisms.



From ftp-wg-owner@hethmon.com  Sun Jan 17 08:47:36 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id IAA01814
	for <ftpext-archive@lists.ietf.org>; Sun, 17 Jan 1999 08:47:35 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Sun, 17 Jan 1999 08:43:28 -0500
Received: from mexnet01.mcsa.net.mx (mexnet01.mcsa.net.mx [200.33.47.216] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Sun, 17 Jan 1999 08:43:26 -0500
Received: from mexpru.mcsa.net.mx ([200.33.47.221]) by mexnet01.mcsa.net.mx (8.8.5/SCO5) with ESMTP id GAA01682; Sat, 16 Jan 1999 06:47:18 -0600 (CST)
Received: from jamie ([204.244.195.142]) by mexpru.mcsa.net.mx (8.8.5/SCO5) with SMTP id FAA20461; Sat, 16 Jan 1999 05:38:46 -0600 (CST)
Message-Id: <199901161138.FAA20461@mexpru.mcsa.net.mx>
Date: Sun, 17 Jan 1999 08:43:26 -0500
X-OldDate:   Sat, 16 Jan 1999 05:38:46 -0600 (CST)
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: 8bfz@mailcity.com
To: friend@exite.com
Subject: Ftp-WG:   Tyson vs Botha Wannna Bet!



 
Feeling Lucky??

CASH IN ON thE FIGHT!!!! - http://209.67.19.99/want_to_bet/

TYSON  V.s.  BOTHA

SATURDAY NIGHT  JAN 16  9pm!

REAL LIVE LAS VEGAS GAMBLING
Place your bet online with up to the minute odds!
Mike Tyson                  -750    
Frans Botha                +500

http://209.67.19.99/want_to_bet/

<a href="http://209.67.19.99/want_to_bet/index.html">CLICK HERE</a> to enter




From ftp-wg-owner@hethmon.com  Mon Jan 18 01:59:09 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id BAA08299
	for <ftpext-archive@lists.ietf.org>; Mon, 18 Jan 1999 01:59:08 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Mon, 18 Jan 1999 01:51:54 -0500
Received: from usilws34.ca-nethaven.com (usilws34.ca-nethaven.com [206.79.17.234] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Mon, 18 Jan 1999 01:51:52 -0500
Received: from blackhawk ([24.3.227.51]) by usilws34.ca-nethaven.com
          (Netscape Messaging Server 3.01)  with SMTP id AAA1020
          for <ftp-wg@hethmon.com>; Mon, 18 Jan 1999 01:32:45 -0500
Message-Id: <4.1.19990118001535.0207df10@192.168.1.18>
X-Sender: mgmail1@192.168.1.18
X-Mailer: QUALCOMM Windows Eudora Pro Version 4.1 
In-Reply-To: <19951.916561531@munnari.OZ.AU>
References: <Your message of "Sun, 17 Jan 1999 00:51:12 MDT."             <4.1.19990116225446.01782920@192.168.1.18>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Mon, 18 Jan 1999 01:51:52 -0500
X-OldDate:   Mon, 18 Jan 1999 00:44:27 -0600
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: "Mike Gleason" <mgleason@ncftp.com>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST "comment" fact

At 07:25 PM 1/17/99 +1100, Robert Elz wrote:

>---> MLST aarnet
>250- Listing aarnet
> Type=OS.unix=slink:aarnet;Size=6;Modify=19951105154900;Perm=;Unique=keVO1+A 
>/aarnet
>250 End

I retrieved the latest -05 draft and I hope to have some modifications
completed to both client and server for further testing.  A quick question
regarding the symbolic link type:  I had been wondering how to include the
"linked to" portion of the symbolic link in the data and I noticed how you
have it embedded after the OS.unix=slink portion.  That seems reasonable,
but how should we handle the case where the destination contains a
semi-colon or space character?  We may need to provide for an escape
character to be able to embed a space or semi-colon.

Another reason we may want to do this is because I can foresee a need for a
"comment" fact.  For example, on MacOS, the user can embed a few lines of
text as a description for a file which could easily have spaces and
semi-colons.  Or, on Windows, a "version" fact which can print the version
resource information embedded in the file.  Come to think of it, a
"comment" fact may be useful all around as an easy way to attach a
description to a file.  Atleast one FTP client (Alex Kunadze's?)  looks for
files named "README" or "CONTENTS" and then in directory listings, try to
match up lines from the downloaded CONTENTS file right along side the
filenames in the listings.  Something like this wouldn't take off and be
useful unless the spec mentioned it as a possible optional fact to implement.

Also, just my personal preference, I like OS.UNIX=symlink better than
OS.unix=slink.  (UNIX since we use UNIX.mode in the spec, and symlink just
because.)

Mike Gleason




From ftp-wg-owner@hethmon.com  Mon Jan 18 02:23:01 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id CAA14712
	for <ftpext-archive@lists.ietf.org>; Mon, 18 Jan 1999 02:23:00 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Mon, 18 Jan 1999 02:18:32 -0500
Received: from munnari.OZ.AU (munnari.oz.au [128.250.1.21] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Mon, 18 Jan 1999 02:18:30 -0500
Received: from mundamutti.cs.mu.OZ.AU ([128.250.1.5])
	by munnari.OZ.AU with SMTP (5.83--+1.3.1+0.57) id HA24875;
	Mon, 18 Jan 1999 18:18:19 +1100 (from kre@munnari.OZ.AU)
In-Reply-To: Your message of "Mon, 18 Jan 1999 01:51:52 CDT."
             <4.1.19990118001535.0207df10@192.168.1.18> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Message-Id: <8365.916643898@munnari.OZ.AU>
Date: Mon, 18 Jan 1999 02:18:30 -0500
X-OldDate:   Mon, 18 Jan 1999 18:18:18 +1100
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Robert Elz <kre@munnari.OZ.AU>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG: MLST "comment" fact 

    Date:        Mon, 18 Jan 1999 01:51:52 -0500
    From:        "Mike Gleason" <mgleason@ncftp.com>
    Message-ID:  <4.1.19990118001535.0207df10@192.168.1.18>

  | At 07:25 PM 1/17/99 +1100, Robert Elz wrote:

Mike is referring to some mail we exchanged - we have been testing
against each other's servers (so, no, the list didn't break for the
rest of you).   And when I sent the MLST data to Mike, the MLST
output fact line was all one line, it hadn't been wrapped (or if it
was, my mailer needs fixing, as that wasn't what was supposed to be sent).

  | That seems reasonable,
  | but how should we handle the case where the destination contains a
  | semi-colon or space character?

Very good question (and a lot of other chars that aren't legal as well).

  | We may need to provide for an escape
  | character to be able to embed a space or semi-colon.

That, or encode the whole thing somehow.   Or don't provide the
target in the MLSx output at all, but make some other way to fetch
that information.

Aside for other readers - this all refers to symlinks that point to
nothing (there is no file at the target location, or the file there
can't be referenced by the authorised user, or various other wacky
error cases).   Where the target exists, the server should refer to
the target for all the facts (including the file type, so the symlink won't
be visible as such - the unique fact will also represent the target file,
so the client will know that a link of some kind was used between the
files).

  | Another reason we may want to do this is because I can foresee a need for a
  | "comment" fact.

Is there support in the group for this to become a standard fact type?
If so, how do we do the encoding?

  | Also, just my personal preference, I like OS.UNIX=symlink better than
  | OS.unix=slink.  (UNIX since we use UNIX.mode in the spec, and symlink just
  | because.)

"UNIX" and "unix" and "Unix" (and all the rest) are the same thing, it is
case independent.   As my server is more for generating examples for the
draft than for production use I'm sometimes weird in not generating what
the draft might suggest is the "normal case" for case independent strings.

As for "slink" - I wimped ... some people like "symbolic link" others like
"soft link", "slink" let me win both ways...

That's the kind of detail (along with how to encode the value) that we
need to sort out when we register a set of UNIX.* OS dependent facts and
file types with the IANA, after the doc itself is published.

Ideally there will be two or three servers around that I can send some
queries at to assist in generating examples for the draft.   Even better
would be if (for now) they did the system dependent things differently,
so I can put in these kinds of examples in the draft without thereby
suggesting what the registered system dependent labels ought to be.
That is, if I can show "OS.UNIX=symlink" and "OS.unix=slink" (and maybe
even more) in adjacent examples, and explain that they show how OS
dependent types can be used - but that the IANA register should be
consulted to obtain the actual types names to use (or to add new ones),
as clearly the inconsistent usages in the examples cannot themselves
be the thing to do.

kre




From ftp-wg-owner@hethmon.com  Tue Jan 19 20:34:12 1999
Received: from mail.hethmon.com (mail.hethmon.com [208.147.156.6])
	by ietf.org (8.8.5/8.8.7a) with SMTP id UAA03567
	for <ftpext-archive@lists.ietf.org>; Tue, 19 Jan 1999 20:34:11 -0500 (EST)
Received: from mail (mail.hethmon.com [208.147.156.6] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Tue, 19 Jan 1999 20:30:25 -0500
Received: from mail.hethmon.com (sambo.hethmon.com [208.147.156.4] ) by mail.hethmon.com
    (Hethmon Brothers Smtpd) ; Tue, 19 Jan 1999 20:30:22 -0500
Message-Id: <199901192030.2225363.6@mail.hethmon.com>
X-Mailer: Post Road Mailer for OS/2 (Green Edition Ver 3.0)
Date: Tue, 19 Jan 1999 20:30:22 -0500
X-OldDate:   Tue, 19 Jan 1999 20:30:11 -0400
Sender: ftp-wg-owner <ftp-wg-owner@hethmon.com>
X-Listname: ftp-wg@hethmon.com
Reply-To: FTPEXT Working Group <ftp-wg@hethmon.com>
From: Paul Hethmon <phethmon@hethmon.com>
To: FTPEXT Working Group <ftp-wg@hethmon.com>
Subject: Ftp-WG:   MLST "comment" fact

** Reply to note from FTPEXT Working Group <ftp-wg@hethmon.com> Mon, 18 Jan 1999 01:51:52 -0500


> I retrieved the latest -05 draft and I hope to have some modifications 
> completed to both client and server for further testing.  A quick question 
> regarding the symbolic link type:  I had been wondering how to include the 
> "linked to" portion of the symbolic link in the data and I noticed how you 
> have it embedded after the OS.unix=slink portion.  That seems reasonable, 
> but how should we handle the case where the destination contains a 
> semi-colon or space character?  We may need to provide for an escape 
> character to be able to embed a space or semi-colon.

My personal preference would be to leave that to the server to escape it. Don't
make it part of the protocol to have to scan every fact line for escape
sequences. None of the facts defined in the draft require spaces or semi-colons
to work. So any fact which is later defined which needs to use spaces or
semi-colons should also specify the escape mechanism for them which works
withing the no-space and no-semi-colon limits. Perhaps using hex notation.

Paul



-------------------------------------------------------------------------
Paul Hethmon                                         phethmon@hethmon.com
Inet.Mail Internet Mail Server                     http://www.hethmon.com
-------------------------------------------------------------------------
Author   "Illustrated Guide to HTTP"   http://www.manning.com/Hethmon?882
-------------------------------------------------------------------------
Warpstock    --    Tune In! & Warp Out!    --    http://www.warpstock.org



