From: Dirk Nimmich (nimmich@muenster.de)
Date: Tue Aug 06 2002 - 17:36:12 CDT
Russ Allbery wrote:
> Decoding is not particularly expensive. I'd be more worried about
> the implementation complexity and the need for a general upgrade
> of news server software than the speed issues.
I can't see any reason why a general upgrade of news servers
because of the encoding would be necessary at all. News servers
create groups with the encoded name, feed groups with the encoded
name, get fed groups with the encoded name, get asked for articles
in the groups with the encoded name, supply articles in groups with
the encoded name, submit articles posted to a moderated group with
an encoded name to the submission address built as usual with dots
replaced by hyphens as the local part. They would never use the
decoded name and never have to reencode the name to submit articles
to a moderation address.
The only software that has to be changed is client software. And
this is only to display an encoded name decoded correctly. Old
client software will see and display the encoded group name,
retrieve postings from the group with the encoded name, display the
encoded group name, post to the group with the encoded name. New
client software will decode the name, display it properly, and
encode it for posting. If it fails to encode the name correctly,
the server rejects the posting because the newsgroup would not
exist.
The only requirements for the encoding are that it must produce a
unique output for a given input (this rules out RFC 2047, if it
would ever be considered) and that the output is a valid group name
in current best practice, i. e. the alphabet for an encoded
component is ALPHA / DIGIT / "+" / "-" / "_". I don't know exactly
how restrictive current server implementations are in this field,
but if plain UTF-8 gets though without problems this can be lowered
a bit so that the only restriction is atext from RFC 2822.
Most of the problems we are and were talking about here is because
there are people on this list who cannot imagine that 1:1 encoding
for transportation is not always the best solution and who mandate
that they must be able to display articles right off the wire
without any processing.
But I fear it's far too late to change this.