From: Bill Davidsen (davidsen@tmr.com)
Date: Fri Aug 20 2004 - 14:32:02 CDT
Frank Ellermann wrote:
> Bill Davidsen wrote:
>
>
>>you open the door to (a) a post with a references header
>>which is not a followup (what, then, is it?
>
>
> It could be the intended begin of a new thread. Or just the
> literal meaning of "reference" (a pointer, See-Also). Or a
> "multi-parented" reply (2822 In-Reply-To).
It could be intended as anything, but software can't tell, therefore I
don't see that we can or should care. If an article with a references
header would not be a followup then we have to name and define what it
is and how to tell. Absent the "how to tell" I return to the conclusion
that we have to treat every article with a References header as a
followup, and thread it if the user desires.
>
>
>>or (b) a followup which does not have a references header.
>
>
> That could arrive at a gateway (web2news, mail2news), and it
> should be handled if possible (2822 algorithm), otherwise it
> would be better to call it "reply" instead of "followup".
How would you tell (in software again) that it was a followup? And how
would you thread it? Since software can't do any of the usual threading
things with the article, what how is it treated.
>
>
>>I find both of those unacceptable, since I can't identify
>>them with software, and therefore clients can't thread them
>>correctly (whatever that means).
>
>
> Using References + Subject should allow to identify threads
> in almost all cases correctly.
I agree, but we can't define how a user agent should do anything, and I
think we should settle for defining a followup as something that can be
unambiguously identified but software in every case. We can denote
common practice of reader agents and "back references," but I think it's
a poor standard to have any question about what is and isn't a followup.
-- bill davidsen <davidsen@tmr.com> CTO TMR Associates, Inc Doing interesting things with small computers since 1979