Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu> writes:
I agree that 2047 is not suitable for this purpose. My guess is that
the IDNA stuff is the best starting point.
IDNA certainly will be the easiest to implement for the servers. If it
catches on in e-mail, I expect it will also be the easiest to implement
for the clients, since they'll have to implement it anyway for e-mail
addresses.
Note that there is no way to avoid a 7-bit encoding for newsgroup names
without significant changes to the Usenet architecture, since posts to
moderated groups are sent to <group-name>@moderators.isc.org by default
and to some variation on that including the group-name even for
hierarchies that don't use the isc.org forwarding system. This requires
embedding the newsgroup name in an SMTP envelope, and I don't expect any
8-bit encoding to work reliably enough there in the near future.
Note that several mail systems are involved, including the mail system at
the originating site, the mail system of the various sites around the
world involved in the moderators.isc.org relay service, and the mail
system of the moderator. Even if we can upgrade the mail systems of the
moderators of non-ASCII groups (likely) and the mail systems of all hosts
volunteering to be part of moderators.isc.org (possible), upgrading the
mail systems of all sites from which posts to a non-ASCII moderated group
may be sent seems pretty intimidating.
Furthermore, it had already been determined that some sort of 7-bit
encoding of the Newsgroups header would be necessary when sending the
messages to the moderators, an encoding that could then be handled by
their software.
Given that a 7-bit encoding is going to be needed *anyway*, I personally
am not particularly fond of the idea of having two separate encodings
floating around, but I can also see the cleanliness argument behind using
pure UTF-8 in the transit system.
--
Russ Allbery (rra(_at_)stanford(_dot_)edu)
<http://www.eyrie.org/~eagle/>