[ietf-nntp] Further syntax
rra at stanford.edu
Fri Mar 5 09:37:50 PST 2004
Charles Lindsey <chl at clerew.man.ac.uk> writes:
> But I am not sure that "clients MUST be prepared" actually means
> anything. AFAICS, the only places where UTF-8 might occur (you listed
> them) are such that the NNTP server itself never needs to look at them
> as part of the protocol. It just get a load of octets from some place (a
> Newsgroups file, for example) and serves them up. So a cooperating
> subnet that chose to use one of those Chinese GB* charsets could still
> use a bog-standard NNTP server already in exact compliance with out
> standard (assuming, that is, that it just took the bunch of octets on
> trust and did not try to use one of those "valid UTF-8 detectors"
Right. That's exactly my concern, although you're mistaken about one
> I think the only place where the internals of an NNTP server _might_
> need to look closely at any UTF-8 would be if USEFOR were to declare
> that newsgroup-names were to be in UTF-8 (Note that USEFOR has currently
> postponed any such decision to a future document).
No, NNTP is already declaring that newsgroup names are UTF-8. We're not
waiting for USEFOR. That's actually one of the goals of our standard,
that all parameters to NNTP commands be UTF-8 (and newsgroup names are the
main ones there).
> So I suspect that the simplest thing for our standard to say is that
> everything MUST be (or be assumed to be) in UTF-8.
Er, I don't see how you can say that given the above. Your first
paragraph is a clear argument against doing that, because people won't
actually make everything UTF-8.
I have no interest in producing a standard people are going to just
ignore. We have to tread the line between doing the right thing and
documenting existing practice here.
> But, strictly speaking, any such on-ASCII encountered on Usenet right
> now is already in breach of some standard, so breaching one more
> standard is neither here nor there. :-(
This is one of the things that I think we should fix, to what degree that
we can, since for as long as what's actually happening on Usenet is
routinely in violation of the standards, the standards are not useful for
their primary purpose (writing interoperable code).
Russ Allbery (rra at stanford.edu) <http://www.eyrie.org/~eagle/>
More information about the ietf-nntp