ietf-nntp OVER and PAT

[Andrew Gierth]

>>>>> "Charles" == Charles Lindsey <chl at clw.cs.man.ac.uk> writes:

 >> If Andrew says doing this causes problems in practice, I'm
 >> inclined to go with that as a good reason not to do things this
 >> way.  It's theoretically cleaner, but theory and practice have had
 >> a long time to diverge.

 Charles> I think we have to ask Andrew to explain his problems again.

There are existing clients that expect the overview (especially the
Subject header) to be the same across multiple servers.

While we've demonstrated that this doesn't happen in the presence of
folding or control characters, at present a non-folded header
containing no tabs or other non-printable characters appears in the
overview _exactly_ as it appears in the article header, and therefore
the same across all servers.

I regard this property (overview exactly equal to raw header, provided
the raw header has no control characters) as being far too important
to mess with. We should of course try and fix what to do about folds,
tabs, nulls, and other non-printing characters, where currently there
is no consistent behaviour, but we shouldn't break things that
currently _are_ consistent across all common implementations.

Bear in mind that a) our spec will not change the world, and b) it is
usually infeasible for OVER and XOVER to behave differently on a given
server (because the munging of overview headers is usually done when
the article arrives).

-- 
Andrew.