From: Thorfinn (thorfinn@tertius.net.au)
Date: Wed Jun 11 2003 - 03:15:35 CDT
On Mon 09 Jun 2003 at 01:21:35PM -0700, in <Pine.LNX.4.53.0306091226180.10379@a>,
John Stanley <stanley@peak.org> wrote:
> On Mon, 9 Jun 2003, Usenet News Support wrote:
> > I know you will continue to misunderstand this, "Re" is not only for the
> > convenience of software, it is useful to the humans using the software.
> I know you will continue to refuse to answer the question, but FOR WHAT
> PURPOSE that is NOT better served by the mandatory References header?
Hrm, yes, scorefiles, definitely, now that someone else posted about it.
It's been in the back of my mind that there *is* a significant
difference between "References: " without "Subject: Re: " and
"References: " *with* "Subject: Re: ", but I'd not been able to put my
finger on exactly it.
I (in an attempt to do something similar to the Forte-Agent user someone
else posted about, which is what crystallised this thought for me) *do*
mildly upscore articles with "Subject: " headers that do *not* begin
with "Re: ", in order to catch new posts *and* thread-mutation in the
same scoring rule.
That's a *lot* easier to do with "Subject: Re: " being well defined.
All the information required to score the article that way is embedded
in the article itself.
I don't need my reading agent to look up the supposedly previous article
in "References: " and check if the subject changed between this article
and the previous one. And even if my reading agent could do that, I
can't guarantee that the previous article exists on my news server, or
will *ever* exist on my news server.
Plus, I can't think of a scoring syntax that encompasses that kind of
cross-article scoring without getting *really* overcomplicated for an
end user. You'd have to go a long way towards having a programming
language, rather than a relatively simple scorefile syntax.
[snippage]
> The truth is, under some fairly common conditions, the References header
> contains ALL of the information that the "Re: " hack contains, plus more,
> and under some fairly common conditions, relying on the "Re: " hack will
> get things completely wrong. And, the truth is, the things that will "work
> better" are things that are USENET-ignorant and have chosen to be that
> way.
Nope. See above.
> > I fail to see what will work better in any way without it.
> Who the fuck keeps talking about doing without it, other than you and
> Charles?
Err, *you*, apparently, at least by implication. If you say:
1. "References: " does everything that "Subject: Re: " does.
2. We MUST NOT say anything about the field text of "Subject: "
that implies that "Subject: Re: " is not necessary (because it's
completely redundant), and should be done away with.
We *are* publishing a document called USEAGE... if we don't put "usage"
things in there, then what on earth point is having it in the first
place?
As far as consensus goes, I haven't bothered chiming in again so far,
because my position has already been stated quite thoroughly once. Not
to mention that we already got told not to repeat our positions over and
over again, since the chairs do actually know how to read and comprehend
email messages, and will determine consensus for themselves.
However, just for the record, that position is now, especially since the
revelation described at the top of this message, as follows:
1. I definitely want USEXXX to require, at least as a SHOULD, the way
that "Subject: Re: " is to be handled by followup agents, and to
require, at least as a "SHOULD NOT automatically generate", stuff like
"Re: Re: " and "Sv: ". The wording that Charles posted a short while
ago, I'd be entirely happy with.
2. I'm entirely happy for the requirement to be in USEAGE, rather than
USEFOR, and am leaning strongly on the side of "Put it in USEAGE".
3. I agree that we should be silent on other agent modifications to
"Subject: ", and leave it to the "market" to decide what they want or
not want to happen there.
4. I think we should be silent on what a Reading Agent (or any other
agent, including gatewaying ones) does with "Subject: Re: ", since it
doesn't matter all that much, and reading agents are generally free to
determine their own functionality. It doesn't terribly harm usenet as a
whole if a specific reading agent is broken. There are plenty of good
alternatives that exist out there.
5. I think that arguing about "oh but it's a magic unstructured header,
so how can we possibily impose requirements on it" is just plain silly.
It's an extremely minor semantic point, and not one that is terribly
relevant. This is a standards document, we can impose any kind of silly
requirements we want. We do want to impose requirements that will be
listened to, but I think that the tiny semantic point of "unstructured
vs structured" headers is not significant to the vast vast majority of
software authors and users. I suspect that most of them just consider
them all as "headers", without bothering to make some magic distinction
between structured and unstructured.
Later,
Thorf
--
<a href="http://tertius.net.au/~thorfinn">thorfinn@tertius.net.au</a>
"Don't think of it as being fired... think of it as leaving early to
avoid the rush."
-- thorfinn^tertius.net.au