Charles Lindsey <chl(_at_)clw(_dot_)cs(_dot_)man(_dot_)ac(_dot_)uk> writes:
That is actually a very good point. There is an expectation, built into
NNTP, that users will be able to filter newsgroup-names using "wildmats"
(which are a restricted form of regular expression). The draft of the
new NNTP standard (draft-ietf-nntpext-base-15.txt) goes to considerable
trouble to define wildmats in such a way that they will work with UTF-8.
Which is an argument in favor of using UTF-8 newsgroup names on the wire
between news servers, since then a UTF-8-aware wildmat will work as one
expects. If newsgroup names are decoded into UTF-8 before matching,
wildmat matches will always work as expected.
I can see why Andrew proposed what he did. Re-encoding at the news/mail
boundary for only those three headers (Newsgroups, Followup-To, and
Control) may be easier than trying to push IDNA all the way down into the
server and dealing with things like wildmat patterns.
--
Russ Allbery (rra(_at_)stanford(_dot_)edu)
<http://www.eyrie.org/~eagle/>