RFC-2047 vs. INN Compatibility Issue

RFC-2047 vs. INN Compatibility Issue

[Matthias Andree]

Hi,

Gnus just generated an article that news.cis.dfn.de aka news.fu-berlin.de
didn't accept (they run some INN version I can't figure out via NNTP as they
put me into NNRPD directly rather than INND). Their server complained 441
437 Missing "Subject" header.

What happened? Gnus generated a header like this in a reply:

Newsgroups: de.comm.software.mailserver
Followup-To: poster
Subject: 
 =?iso-8859-15?q?Re:_kein_Anmeldung_mit_pop3_m=F6glich_bei_courier-imap?=

Looks like this is fine according to RFC-2822 and 2047, but it is not a
recommended way to encode the header, rfc2047.el encodes too much currently.

Anything before the first word that needs encoding and anything after the
last word that needs encoding should be left unencoded, at least and
particularly the Re: should never be part of an encoded string. Rationale:
non-MIME-Aware readers will duplicate the Re: if it's encoded like this.

I'd suggest that the header be written as (I don't care about the
indentation here):

Subject:·Re:
 =?iso-8859-15?q?_kein_Anmeldung_mit_pop3_m=F6glich_bei_courier-imap?=

or even better:

Subject: Re: kein Anmeldung mit pop3 =?iso-8859-15?q?m=F6glich?=
 bei courier-imap

Care must be taken not to encode each word individually because that's
inefficient, and because adjacent encoded-words get their separating LWS
_removed_, so =?iso-8859-1?q?test?= =?iso-8859-1?q?word?= becomes
"testword", not "test word".

Can anyone help with a rfc2047.el patch?

Thanks in advance,
Matthias

Re: RFC-2047 vs. INN Compatibility Issue

[Russ Allbery]

Matthias Andree <matthias.andree@gmx.de> writes:

> Gnus just generated an article that news.cis.dfn.de aka
> news.fu-berlin.de didn't accept (they run some INN version I can't
> figure out via NNTP as they put me into NNRPD directly rather than
> INND). Their server complained 441 437 Missing "Subject" header.

> What happened? Gnus generated a header like this in a reply:

> Newsgroups: de.comm.software.mailserver
> Followup-To: poster
> Subject: 
>  =?iso-8859-15?q?Re:_kein_Anmeldung_mit_pop3_m=F6glich_bei_courier-imap?=

> Looks like this is fine according to RFC-2822 and 2047, but it is not a
> recommended way to encode the header, rfc2047.el encodes too much
> currently.

All versions of INN will reject that.  (It's a long-standing bug in the
INN header parsing.)

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: RFC-2047 vs. INN Compatibility Issue

[Matthias Andree]

Russ Allbery <rra@stanford.edu> writes:

> All versions of INN will reject that.  (It's a long-standing bug in the
> INN header parsing.)

Heiko Schlichting says this bug has been fixed on 2001-04-23 (at least
for the DEVEL version). I haven't checked it myself, apparently my 2.3.X
version deosn't accept this header.

At least news.cis.dfn.de will accept that now (they fixed theirs), how
far such articles propagate is a different story though.

Does this warrant a workaround in Gnus?

-- 
Matthias Andree

Re: RFC-2047 vs. INN Compatibility Issue

[Russ Allbery]

Matthias Andree <ma@dt.e-technik.uni-dortmund.de> writes:
> Russ Allbery <rra@stanford.edu> writes:

>> All versions of INN will reject that.  (It's a long-standing bug in the
>> INN header parsing.)

> Heiko Schlichting says this bug has been fixed on 2001-04-23 (at least
> for the DEVEL version). I haven't checked it myself, apparently my 2.3.X
> version deosn't accept this header.

> At least news.cis.dfn.de will accept that now (they fixed theirs), how
> far such articles propagate is a different story though.

> Does this warrant a workaround in Gnus?

It may now be accepted on transit, but I thought it still wasn't accepted
for new posts.  But I could be wrong; I haven't checked.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: RFC-2047 vs. INN Compatibility Issue

[Dave Love]

Russ Allbery <rra@stanford.edu> writes:

> All versions of INN will reject that.  (It's a long-standing bug in the
> INN header parsing.)

What exactly is the bug (so I can make sure it is avoided)?

Re: RFC-2047 vs. INN Compatibility Issue

[Russ Allbery]

Dave Love <d.love@dl.ac.uk> writes:
> Russ Allbery <rra@stanford.edu> writes:

>> All versions of INN will reject that.  (It's a long-standing bug in the
>> INN header parsing.)

> What exactly is the bug (so I can make sure it is avoided)?

The first line of the header has to contain some non-whitespace content,
even if it's folded.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: RFC-2047 vs. INN Compatibility Issue

[Dave Love]

Matthias Andree <matthias.andree@gmx.de> writes:

> particularly the Re: should never be part of an encoded string. Rationale:
> non-MIME-Aware readers will duplicate the Re: if it's encoded like
> this.

That's a good point.  I'll change it.  Note that the CVS version has a
fix which should avoid the break after `Subject:'.