ietf-nntp Simplification of wildmat

[Russ Allbery]

Okay, I've read all of the wildmat thread (well, skimmed it this time
around).  After thinking about this for a while, and particularly given
that we've now dumped XPAT back to the status of a non-standardized
command, I think Andrew is right.

Furthermore, I think that introduction of character sets also has no
motivating application for the base draft, and therefore we may as well
leave them out as well.  This has the advantage of considerably
simplifying this entire section, giving the most freedom to implementors,
and allowing the extended wildmat syntax to be advertised as an extension.

Accordingly, I'm wondering if we shouldn't just start over again from
first principles, ban the characters '!', '[', ']', ',', and '\' from
wildmats entirely, and specify a wildmat language that involves only UTF-8
characters matching themselves, *, and ?.

NEWNEWS can get a simple explanation of commas and ! to negate a pattern,
possibly taken from Clive's proposed text, and have that be unique to
NEWNEWS.

All of this work in describing a more complicated wildmat syntax can then
move into a separate document if need be and be advertised as an
extension, and the presence of that extension in LIST EXTENSIONS will
indicate that all wildmats will be interpreted using the extended syntax,
which includes , as part of the language.

Since the extra special characters are banned from wildmats otherwise,
clients SHOULD NOT send them unless the server advertises this extension,
which will also mean that backward-compatibility is preserved; servers
using the extended syntax by default won't surprise any clients that
follow the protocol.

What do people think of taking that approach?  I think the massive
reduction of complexity in the draft over the amount of language that it's
taking to describe wildmats at present makes it worthwhile, even if it
leaves NEWNEWS different than some other obstensibly similar commands in
the absence of an advertised extension.

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