From: Russ Allbery (rra@stanford.edu)
Date: Mon Jul 02 2001 - 18:07:56 CDT
Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
> Yes, but in this particular case all the various encodings are a "MUST
> accept" for all agents, so the implementation effort required with
> either wording is much the same.
No, this is not correct. If we simply allow MIME, then one can use
existing MIME implementations and we know exactly how they work and what
the issues are from dealing with them in mail. There is therefore no
additional testing requirements *specific to this draft*, only testing
requirements that apply to any MIME-compliant software, including mail
readers. This means that no additional *protocol* testing would really be
required, as MIME is a well-understood and solved problem.
> But the wording is written, so all the exceptions are there to be seen.
Yes, that's why I'm providing an alternative wording that I think is
better. The wording currently in the draft I consider to be unacceptable.
> The only practical difference between them is the pressure that is
> applied to the writers of posting agents.
The practical difference is extreme and unnecessary complication of our
specification, plus all of the other problems that I spelled out in my
previous messages.
>> It's at least more important than the Archive header; if I had the
>> choice, I'd rather spend my time arguing about this.
> But I see no need to have the argument,
I do.
> because the effect on what implementors actually have to DO is rather
> small
The effect on what the implementors have to read and understand is huge,
as are the implications for being able to reuse code already written for
mail.
>> I vote that we decide this is out of scope and will just needlessly
>> complicate our standard, given that we don't anticipate widespread use
>> of message/partial on Usenet anyway, and at most note this problem in a
>> NOTE. Those who care can take up the matter with ietf-822 and hammer
>> out a better definition in a future revision of the MIME standards.
>> References is now a mail header, so it's well within scope for such a
>> revision.
> No, References is primarly is News header, which mail has now adopted
> (but which is not all that regularly used).
Wrong. I'm on innumerable mailing lists and read my mail with a threading
mail reader, and something like 90% of the mail messages that I get use
References correctly.
> Mail user agents mostly do not implement threading
Wrong. Even Pine implements threading now.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>