Re: [PATCH] chomp to remove '\r' as well

[Joel Aelwyn <joel@lightbearer.com>]

[-- Attachment #1: Type: text/plain, Size: 2330 bytes --]

On Mon September 11 2006 07:18, Mads Martin Joergensen Mads Martin 
Joergensen <mmj@mmj.dk> wrote:
> * Joel Aelwyn <joel@lightbearer.com> [Sep 11. 2006 15:12]:
> > > Anyone object to this behaviour? Just heard on irc someone got
> > > bitten by this:
> > >
> > > --- chomp.c     20 Feb 2006 05:00:09 -0000      1.4
> > > +++ chomp.c     11 Sep 2006 12:34:28 -0000
> > > @@ -38,7 +38,7 @@
> > >
> > >         i = strlen(str) - 1;
> > >
> > > -       while(str[i] == '\n') {
> > > +       while(str[i] == '\n' || str[i] == '\r') {
> > >                 str[i] = 0;
> > >                 i--;
> > >         }
> >
> > As far as I recall, getting anything with \r\n as an EOL pretty
> > much grossly violates the SMTP RFCs, so I'll make two suggestions
> > about this:
> >
> > 1) Limit it *strictly* to a single \r\n, not any random combination
> > of the two characters. You'll drastically reduce your chances of
> > subtle and nasty misbehaviours being introduced when (not if) some
> > software comes along and relies on things obeying the RFCs.
> >
> > 2) Make it a configurable option. Call it 'broken-CRLF-support' or
> > somesuch, because it most assuredly is.
> >
> > Outlook is one of the primary offenders in this arena, for what
> > it's worth; DJB had -- probably has -- a rant up about it and why
> > qmail does/did not support a similar feature (it's been years since
> > I used qmail, but the combination of the two provided no end of
> > headaches back when I had to deal with them both at my workplace...
> > and, to a one, they were caused by Outlook blatantly violating the
> > protocol specifications.
>
> Ah, this has nothing to do with the smtp layer. SMTP always have to
> have \r\n, and that's handled in send.
>
> Problem is that when I for instance read a subscriber from a
> subscribers file with the old behaviour and match
>
> "mmj@mmj.dk" with "mmj@mmj.dk^M", they're different.
>
> Same thing when reading control files.

Ah, ick. But, in that case, being fancy is probably not warranted... 
anyone who would get bitten by the broader behaviour is already doing 
something they need to be smacked for. Comments withdrawn, or at least 
deferred into the 'nice idea, maybe, but then again maybe not' 
bucket. :)
-- 
Joel Aelwyn <joel@lightbearer.com>

[-- Attachment #2: Type: application/pgp-signature, Size: 191 bytes --]