From: Bruce Lilly (blilly@erols.com)
Date: Sun Feb 16 2003 - 16:47:18 CST
J.B. Moreno wrote:
> Indicating future directions is IMO fairly useless unless you are also
> /starting/ down that path -- which means generation as well as acceptance.
Consider RFC 2822. It introduced parse-only rules for recognizing
unquoted dot in a phrase, e.g.
From: J.B. Moreno <planb@newsreaders.com>
instead of the standards-compliant
From: "J.B. Moreno" <planb@newsreaders.com>
Unquoted dot has to date been illegal in a phrase, as
the dot is a special. But by first requiring recognition
while still prohibiting generation, the way is being
prepared to legalize that contruct. Generating now
would be a mistake as many parsers are not yet prepared
to accept what has always been an illegal construct.
> The existing infrastructure *already* has to deal with it, because the
> existing infrastructure is encountering it somewhere every second of every
> day.
Really? Name three moderated newsgroups with non-ASCII
names. Name three implementations of injection agents
that transform raw utf-8 in header fields to RFC 2047,
preserving information about the originator's (untagged)
charset and language, before mailing to moderators. Name
three news-to-mail gateways that do so.
> Now, some of the infrastructure may not be dealing with it in the best
> possible way (IMAP and SMTP dropping bytes), but *is* happening.
>
> (And by the way, if other protocols get to criticize us, I think I should
> get to criticize them, and dropping data just because you don't know what to
> do with it is *stupid*).
The criticism is that the draft (N.B. drafts are not
standards) proposes to break backward compatibility
with existing, widely-implemented protocols that are
in full conformance with RFC 1036, for no good reason
(i.e. there are backwards-compatible mechanisms
available). Dropping *illegal* data which has always
been illegal data is perfectly acceptable. Forcing
illegal data on protocols is not acceptable. Your
criticism should be directed towards the software
(primarily news-specific UA) authors who have not
implemented MIME, not towards authors of Standards
Track RFCs documenting protocols which are in full
compliance with IETF guidelines and Best Current
Practice.
If you have a complaint about the dropping of *legal*
data, that's another matter. If so, let's hear it.
> A blessed charset *will* work -- in fact that's how most of news currently
> handles the situation, we want to define a /new/ blessed charset, one that
> is universal instead of a group or hierarchy charset.
What evidence do you have that Chinese (mainland and
Taiwan) users will do anything differently from what
they're doing now? Or for that matter, the Koreans,
particularly in light of the ISO 10646 Amendment 5
fiasco? Or the Russians, or... You're claiming
that it will work; let's see some evidence. Let's
see if you can convince the cn hierarchy to switch
to utf-8. It should be an interesting experiment.
> The people managing groups (whether officially or through being a netcop)
> generally care enough to follow official standards and to put pressure on
> others to do the same -- the fact that they aren't doing so now is a clear
> indications that the standards have failed to keep up with reality.
No it's a crystal clear indication that you're wrong
about them applying pressure for standards compliance.