A T more intelligent than ^ ?

A T more intelligent than ^ ?

[Ami Fischman]

If an article ("artB") is a reply/followup to another article ("artA"),
and artB is currently displayed then hitting `^' in the summary buffer
usually brings up artA.  However, if artB was created by a brain-dead
mailer, it might not have the References: header.  Using ^ in this case gets
on a message that the article has no References: and no parent.  Using 
`A T', however, gnus picks up the parent just fine (presumabely it does so
by examining the subject lines and figuring things out for itself).  Can we
have ^ fall back to the `A T' logic if no References: are found?

FWIW, by ^ I'm referring to gnus-summary-refer-parent-article and by `A T'
I'm referring to gnus-summary-refer-thread.

Cheers,
-- 
  Ami Fischman
  usenet@fischman.org

Re: A T more intelligent than ^ ?

[Lars Magne Ingebrigtsen]

Ami Fischman <usenet@fischman.org> writes:

> Using `A T', however, gnus picks up the parent just fine
> (presumabely it does so by examining the subject lines and figuring
> things out for itself).

Not quite -- `A T' uses the normal Gnus threading code, which
examines References (and if you're not using NOV) In-Reply-To.  `^'
also does this, but References and/or In-Reply-To may be broken in
interesting ways, so it is possible that `A T' gives different
results than `^'.  Sometimes better, sometimes worse, I'd have
thought. 

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: A T more intelligent than ^ ?

[Ami Fischman]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

[...]

> Not quite -- `A T' uses the normal Gnus threading code, which
> examines References (and if you're not using NOV) In-Reply-To.  `^'
> also does this, but References and/or In-Reply-To may be broken in
> interesting ways, so it is possible that `A T' gives different
> results than `^'.  Sometimes better, sometimes worse, I'd have
> thought. 

The message I'm using (the followup with missing References header) has no
In-Reply-To header either, so A T must also be looking at the subject header
to determine threading.  But, given that sometimes `^' just gives up,
couldn't we have it at least attempt to pick up the nearest article that A T
finds?  Seems like a logical fallback to me...

To emphasize, neither In-Reply-To nor References appears in the headers of
the followup at all.  The complete list of headers (after gnus pulled it in)
are: X-From-Line, Return-Path, Delivered-To, Received, X-Gnus-Mail-Source,
Message-Id, X-Mailer, Date, From, To, Subject, Mime-Version, Content-Type,
Content-Transfer-Encoding, Content-Disposition, Lines, and Xref.  Nothing
about what this message is a reply or followup to...

Cheers,
-- 
  Ami Fischman
  usenet@fischman.org

Re: A T more intelligent than ^ ?

[Lars Magne Ingebrigtsen]

Ami Fischman <usenet@fischman.org> writes:

> To emphasize, neither In-Reply-To nor References appears in the headers of
> the followup at all.

Hm.  And yet `A T' managed to thread the message?  That's really
odd -- it really shouldn't be able to fetch anything at all.

It's possible (although extremely unlikely) that the news server is
returning a References header in the NOV headers, and not in the HEAD
headers, but I've never heard of anything like that....  And without
References/In-Reply-To, Gnus just can't thread.  (It can gather
articles based on Subject, but that doesn't help `A T'.)

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: A T more intelligent than ^ ?

[Ami Fischman]

This is in an nnml group, not nntp.  No news server in the picture.  Could
it be that Gnus, in building the summary buffer, examines subject headers,
and that then A T uses this info somehow?

Cheers,
-- 
  Ami Fischman
  usenet@fischman.org

Re: A T more intelligent than ^ ?

[David Abrahams]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Ami Fischman <usenet@fischman.org> writes:
>
>> To emphasize, neither In-Reply-To nor References appears in the headers of
>> the followup at all.
>
> Hm.  And yet `A T' managed to thread the message?  That's really
> odd -- it really shouldn't be able to fetch anything at all.
>
> It's possible (although extremely unlikely) that the news server is
> returning a References header in the NOV headers, and not in the HEAD
> headers, but I've never heard of anything like that....  And without
> References/In-Reply-To, Gnus just can't thread.  (It can gather
> articles based on Subject, but that doesn't help `A T'.)

I've very commonly seen the opposite: `A T' finds no messages, but I
can go backwards in time with `^'.

Inexplicable?
-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com

Re: A T more intelligent than ^ ?

[Lars Magne Ingebrigtsen]

Ami Fischman <usenet@fischman.org> writes:

> This is in an nnml group, not nntp.  No news server in the picture.  Could
> it be that Gnus, in building the summary buffer, examines subject headers,
> and that then A T uses this info somehow?

Gnus doesn't use the Subject when threading.  It does use that when
gathering threads -- is that what you mean?  If `A T' gives you
something like this:

R +[  12: Ami Fischman           ] Re: A T more intelligent than ^ ?
R +    <  27: David Abrahams         > 

Then that's not due to `A T' being better at threading than `^', but
just because `A T' fetches a gazillion headers and then gathers based
on Subject.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: A T more intelligent than ^ ?

[Lars Magne Ingebrigtsen]

David Abrahams <dave@boost-consulting.com> writes:

> I've very commonly seen the opposite: `A T' finds no messages, but I
> can go backwards in time with `^'.
>
> Inexplicable?

That can happen if `^' pulls in articles from a different group than
the current group.  If that's not what happens, then it must be a bug
of some kind.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: A T more intelligent than ^ ?

[Ami Fischman]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Then that's not due to `A T' being better at threading than `^', but
> just because `A T' fetches a gazillion headers and then gathers based
> on Subject.

Okay, then, can ^ be modified to fetch the gazillion headers, gather on
subject, and then give the preceding article in the resulting tree?

Or do you prefer to keep the distinction between the underlying behaviour
and prefer that people hit the A T if ^ gets them nothing?

-- 
  Ami Fischman
  usenet@fischman.org

Re: A T more intelligent than ^ ?

[Lars Magne Ingebrigtsen]

Ami Fischman <usenet@fischman.org> writes:

> Okay, then, can ^ be modified to fetch the gazillion headers, gather on
> subject, and then give the preceding article in the resulting tree?

`^' does one thing, and only one thing: It fetches an article based
on Message-ID.  That article may come from another group, another
server or another space-time continuum altogether.

`A T', on the other hand, fetches (bits of) the current group and
tries to see how the current message fits into that.  So...

> Or do you prefer to keep the distinction between the underlying behaviour
> and prefer that people hit the A T if ^ gets them nothing?

... yes.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen