From: Eivind Tagseth (eivindt@multinet.no)
Date: Sun May 02 2004 - 04:04:06 CDT
* Seth Breidbart <sethb@panix.com> [2004-05-02 02:42:30 -0400]:
> 1. As a _reader_ of Usenet, I prefer to see Re: on followups. This
> sometimes enables me to more easily find thread beginnings. The
> References header doesn't do as well; first, topic changes aren't
> picked up (they are imperfectly picked up by Subject changes, but
> that's still better than never). Second, broken posting agents
> often generate followups (in the sense that people think of them:
> articles beginning with an attribution line, quoting a previous
> article, etc.) which lack the References header.
Your second point is the only "valid" argument for the "Re: " prefix as
far as I can see. Why wouldn't topic changes be perfectly picked up by
subject changes (when the followup-agents are compliant)? And what's
to keep a user from changing the subject of a thread, but still keeping
the "Re: " prefix? It's just an abbreviation meaning "in the matter of",
why shouldn't a user manually enter this prefix starting a new subthread?
I can't see how a reading agent would just look at the first 4 letters of
a subject to decide if it is a subject change or not, it just seems to
fragile.
I _can_ see how it is useful for dealing with broken agents though, but
if they are broken enough to omit the references header, then they are
broken enough to use a different "back-reference" than "Re: ", aren't they?
> 2. If the parent article is not available (the likelihood of which
> depends on the site the reader is using), presence or absence of
> Re: tells whether the Subject header changed (imperfectly, but with
> high correlation with accuracy). Therefore, automatically using
> Re: sometimes provides information that would not otherwise be
> available.
I don't get this. If none of the ancestors are available, then yes, it
would be impossible to tell if the subject ever changed. But then again,
who cares? The only way to present it to the user would be as a standalone
thread, with missing parents.
If only the immediate ancestors are missing, you'd still be able to detect
a subject change between the last available article and this one. You wouldn't
be able to tell if the change happened just now, or if one of the missing
ancestors changed it, but how important could that be? If the new subject
was interesting, you'd still read this article, even if the subject actually
changed in a missing article, wouldn't you?
> 3. A reading agent *cannot* perfectly predict whether an article is a
> followup with unchanged header, hence it cannot generate a display
> with Re: precisely when a followup agent would have defaulted to
> having it.
You're saying that it cannot automatically prepend a "Re: " before displaying
the subject to the user, the same way that you could if the followup-agent
automatically prepended it? Yes, that is true (although the followup-agent
may also screw up, as Bruce has illustrated several times). And as Bruce
has shown, you cannot perfectly replace a "Re: " with an "Ant: " for display.
> I have not seen any rationale for disparaging "Re: " other than bogus
> arguments like "Subject is unstructured and the user can do anything
> he wants therefore we shouldn't default in the most useful way."
o It _is_ a hack, and if news was specified from fresh today, I don't think
the "Re: " hack would have been reintroduced.
o The "Re: " hack causes a lot of problems today, with people using
all sorts of variants of "Re: ".
o "In re" is mostly used in english-speaking contries, other countries
will want a localized variant. The "Re: " prefix cannot be perfectly
translated to another language as it is impossible for the reading
agent to know for certain that "Re: " really is a back-reference at
all times.
o As you say yourself above, you prefer to see the "Re: " prefix on followups
because it makes things easier for you to see. To me, this is yet another
indication that the "Re: " prefix does not belong in USEFOR, it is not
part of the protocol, it's a mechanism more intended for humans than
for usenet operations, and thus something that should be dealt with in
USEAGE.
o As for the structured vs. unstructured discussion, I really don't care that
much about how other RFCs have defined the Subject header. If imposing
structure on an unstructured header could fix this problem in a clean
manner, then that could be a good solution. But the "Re: " hack doesn't
come close to a clean solution at all! It's just a prefix prepended to an
unknown text, which may already start with such a prefix and which cannot
with 100% certainty be recognized as a prefix be the reading agent. It's
just plain ugly. If it's really needed, I'd say we put it in a different
header.
I'm sure there are more pros and cons. This discussion has lasted for so
long that it would be quite nice to have them collected.
Eivind