headers
Charles Lindsey | 2 May 2005 12:49
Picon
Picon

Re: Suggested References texts


In <87vf651rad.fsf <at> windlord.stanford.edu> Russ Allbery <rra <at> stanford.edu> writes:

>Charles Lindsey <chl <at> clerew.man.ac.uk> writes:
>> Russ Allbery <rra <at> stanford.edu> writes:

>>> I *don't* want to give the impression, even in USEFOR, that all the
>>> rules in RFC 2822 for replies apply unmodified to followups.  Replies
>>> and followups *aren't* the same thing, and I think it's worth making
>>> that clear.  Among other things, correctly constructing a followup
>>> requires dealing with fields that aren't in RFC 2822 at all.

>> Eh? Which fields might those be?

>Followup-To and Newsgroups.

Ah! My bad. I was thinking just of the References header, not of the other
things a followup agent has to get right.

--

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133   Web: http://www.cs.man.ac.uk/~chl
Email: chl <at> clerew.man.ac.uk      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9      Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5
Permalink | Reply | 0 comments |
headers
Charles Lindsey | 2 May 2005 12:48
Picon
Picon

Re: 2045 ABNF (was: Again broken)


In <4272F833.324 <at> xyzzy.claranet.de> Frank Ellermann <nobody <at> xyzzy.claranet.de> writes:

>Charles Lindsey wrote:

>>> So far I didn't know that usefor-03 implicitly includes
>>> all 2045 terms.  I only knew that it includes 2822 minus
>>> obs, and of course 2234bis.

>> It needs to be stated somewhere.

>It is, I was blind.

Not in the current USEFOR AFAICS.

>  But we cannot use this pseudo-ABNF, it has
>to be fixed for the three (?) 2045 terms needed in usefor-xx:

>value, token, parameter, anything else ?

>> It even appeared (but not normatively) in my old draft-13.

>Yes, the "collected ABNF" was nice,

Actually, it was section 2.4.2 of the old draft-13 I was referring to.

>1) value = token / quoted-string

>2) parameter = attribute "=" value
(Continue reading)

Permalink | Reply | 9 comments |
headers
Charles Lindsey | 2 May 2005 13:49
Picon
Picon

Re: Again broken


In <4272652A.9030101 <at> epix.net> "Forrest J. Cavalier III" <mibsoft <at> epix.net> writes:

>Charles Lindsey wrote:
>> If the WG wants this change, then it ought to be discussed first.

>There you go again.

>Charles, the only meaning of "change" that makes sense is a change
>from current published standards and practice. I believe you have once
>again used it to mean "an edit to the draft."  Please stop it.

The "change" I am talking about is a change from previous WG decisions.

The WG decided years and years and years ago (long before I became editor)
to allow comments in headers (subject to "do not generate yet"). Later on,
it decided to exempt certain headers from this on performance grounds. The
list included Newsgroups, Path, Followup-To, Distribution and Message-ID,
but _never_ Control. All this was reflected in our drafts up to draft-13.

>If something appeared in the draft that was never discussed in the WG,
>it ought to be removed.

Any that is exactly what I am complaining about. The new USEFOR drafts
have changed the syntax of the Control header, contrary to what had been
discussed and incorporated in the earlier drafts, and without any WG
decision to do so.

If the WG now wishes to reopen and discuss this case again, then fine. But
without such discussion, it should remain as it always has been.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Charles Lindsey | 2 May 2005 13:30
Picon
Picon

Re: Obsolete vs. obsolescent


In <87r7gt1r37.fsf <at> windlord.stanford.edu> Russ Allbery <rra <at> stanford.edu> writes:

>Charles Lindsey <chl <at> clerew.man.ac.uk> writes:

>> But evidently Frank doesn't want that to be said (he is complaining
>> about current systems that object when you try to do it), and I don't
>> think I want to do it either. Better just to let the usage die away
>> (which USEPRO requires).

>That isn't going to actually happen in any sort of reasonable time frame.

>> Otherwise we shall find new software coming on stream that objects to
>> 'cmsg' in Subjects (just to comply with the new USEFOR) just as the last
>> remaining servers that might have interpreted that 'cmsg' are being
>> withdrawn.

>I wonder if any of us will live long enough to see the last remaining
>servers that interpreted cmsg be withdrawn.

I hope to, and I am older than you are :-) . If, on an occasional server,
some group gets wrongly created/removed or some article gets cancelled, I
can live with that. It happens all the time, and the effect would not
propagate outside of that server. That hardly amounts to "huge
interoperability problems at the present time".

>> How much of a problem are such headers causing at the present time?

>They aren't propagated.  They have huge interoperability problems at the
>present time.
(Continue reading)

Permalink | Reply | 37 comments |
headers
Charles Lindsey | 2 May 2005 13:18
Picon
Picon

Re: Suggested References texts


In <075201c54d54$d7b2d460$0b01a8c0 <at> isolution.nl> "Ruud H.G. van Tol" <rvtol <at> isolution.nl> writes:

>Charles Lindsey:

>>    references      =  "References:" SP [CFWS] msg-id 1*(CFWS msg-id )
>>                       [CFWS] CRLF

>I suppose that was meant as

>   references      =  "References:" SP [CFWS] msg-id *( CFWS msg-id )
>                      [CFWS] CRLF

>(changed the '1*' into '*'; inserted a space)

Oops!

--

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133   Web: http://www.cs.man.ac.uk/~chl
Email: chl <at> clerew.man.ac.uk      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9      Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5
Permalink | Reply | 0 comments |
headers
Charles Lindsey | 2 May 2005 14:03
Picon
Picon

Re: Again broken


In <4272E032.4BCB <at> xyzzy.claranet.de> Frank Ellermann <nobody <at> xyzzy.claranet.de> writes:

>Charles Lindsey wrote:

>> OK, if you don't want it fixed, let it go.

>I've proposed many ways to fix it, e.g. by clear names for
>the msg-id productions, or the last proposal was this note:

Changing the names of things can drop hints, but it can never change a
normative definition.

>| Note: id-right is a domain, and no-fold-literal is the
>| special case of a domain-literal (a.k.a. address-literal,
>| see RfC 2821)

I don't think USEFOR should be relying on RFC 2821 (RFC 2822 is fine, of
course).

By all means say that <no-fold-literal> is to be treated in some sense
like a <domain-literal>. But that is precisely what my text was trying to
do, but you objected to my text :-( . Just to remind you, my exact words
were:

|    "... any <dot-atom-text> or
|    <no-fold-literal> used for the <id-right> are to be interpreted as
|    <domain>s as described in section 3.4.1 of [RFC 2822].

That covers all you have been asking for, and ensures that the
(Continue reading)

Permalink | Reply | 1 comment |
headers
Charles Lindsey | 2 May 2005 13:14
Picon
Picon

Re: Suggested References texts


In <4272D4E0.3BB8 <at> xyzzy.claranet.de> Frank Ellermann <nobody <at> xyzzy.claranet.de> writes:

>Charles Lindsey wrote:

>> suppose you are a multipart FAQ writer trying to use
>> References to refer back to the earlier parts, how do
>> you know what you can do?

>There are many possible and correct ways to do this, and
>they are all obvious.  It's 2005.  If somebody posts a
>multipart FAQ in an NG where I don't like it I'm tempted
>to report it as net abuse.

Yes, all that you say is true. "Everone" (TINE) "knows" how the References
header is supposed to work. That is, by definition, how "folklore" works.
But Usenet has been running on folklore ever since RFC 1036 was written,
and the whole purpose of our drafts is to codify it properly, because in
many cases different people have understood the folklore in different
ways. The whole purpose of having standards is to prevent that sort of
thing from happening.

Therefore, USEFOR and USEPRO between them MUST document how the References
header is supposed to work (I think we are now agreed that it differs
sufficiently from References in RFC 2822 that we cannot rely on RFC 2822
to any great extent).

Hence why I want to say, in USEFOR, what a well-formed References header
looks like, and what it means (being careful to say no more than is
necessary for current practice, so as not to prejudice future
(Continue reading)

Permalink | Reply | 1 comment |
headers
Russ Allbery | 2 May 2005 18:38
Picon
Favicon
Gravatar

Re: Obsolete vs. obsolescent


Charles Lindsey <chl <at> clerew.man.ac.uk> writes:

> I hope to, and I am older than you are :-) . If, on an occasional
> server, some group gets wrongly created/removed or some article gets
> cancelled, I can live with that.

That's not the real issue.  The issue is that such articles disappear,
sometimes silently, sometimes noisily.  That is the very definition of an
interoperability problem.

> _Some_ servers decline to propagate them. Some don't. Can you quote
> examples that have actually happener to justify that "huge
> interoperability problems" claim?

Haven't we done this multiple times before?  I feel like I'm repeating
myself.  I could have sworn that we'd reached consensus on this repeatedly
in the past, and yet, here we go again.

INN 1.7 is still widely deployed and even recommended for new
installations by, e.g., Debian because it's simple.  Many existing filters
look for "Subject: cmsg" when making filtering decisions as well as
looking at Control headers, so as not to miss articles sneaking in via
older interpretations of the standard.  The change in INN was made in 2.3,
so even for people who want to run INN 2.x, it was made as part of a
significant upgrade that requires a complete overview rebuild, and a fair
number of people have not bothered.

And that's just in the INN world, let alone all of the other news servers
out there.
(Continue reading)

Permalink | Reply | 36 comments |
headers
Charles Lindsey | 2 May 2005 18:37
Picon
Picon

Re: CFWS (was: Again broken)


In <4272DB3E.74CA <at> xyzzy.claranet.de> Frank Ellermann <nobody <at> xyzzy.claranet.de> writes:

>Charles Lindsey wrote:

>> it ought to be discussed

>Kill all CFWS in pre-2045 news-only-headers where it doesn't
>contradict 2045 / 2822.  For new headers like User-Agent or
>Injection-Info adopt common practice or 2045 / 2822 style.

>I know that it's not what the WG once wanted, but you asked.

OK, that's a viable point of view, which the WG might accept.

The headers affected would be:
    Newsgroups
    Distribution
    Followup-To
which we have already decided to make FWS-only. plus:
    Expires
    Approved
    Control
    Supersedes
    Xref
    Lines
and also
    Message-ID
which we decided to make FWS-only, even though it is not a
news-only-header, but not References (though we say "must not generate
(Continue reading)

Permalink | Reply | 16 comments |
headers
Frank Ellermann | 3 May 2005 04:36
Picon
Picon

Re: Suggested References texts


Charles Lindsey wrote:

> Hence why I want to say, in USEFOR, what a well-formed
> References header looks like,

It look likes a msg-id-list.  That's all about its form,
and that's all for USEFOR, the same look and feel as the
RfC 2822 References.

> and what it means

If has no clear meaning, we're only sure that an article
without References is no "followup".  By elementary logic
we found that a "followup" has References.

> A. By stating what properties a well-formed References
>    header should possess.

It should possess the property of a msg-id-list.  If you
want to go into details state that a msg-id should not
occur more than once, and that the Message-ID: msg-id
should not be listed in this msg-id-list.  But better
keep it as simple and stupid as it is => no details.

> B. By setting forth an algorithm for generating it.

Yes, that's what you have in USEPRO, and similar in 2822.
"Reply" in 2822 is a simpler construct than a "followup" -
no Followup-To: in RfC 2822 - but essentially the same.
(Continue reading)

Permalink | Reply | 0 comments |

Gmane