chasing mail threads

chasing mail threads

[Dave Abrahams]

I need to keep my Inbox as empty as possible (sometimes I use
mail clients that don't hide read messages).

I also very commonly need to get back to look at the history of a thread.
Gnus has `A T' and `^' which I use all the time...

but it doesn't work so well in my Inbox because the messages have
usually been deleted.

I have an Archive folder that keeps all the mail I've ever sent or
received, though.  Is there any way to tell Gnus to find my thread
there?

Thanks,

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: chasing mail threads

[Andreas Schwab]

Dave Abrahams <dave@boostpro.com> writes:

> I have an Archive folder that keeps all the mail I've ever sent or
> received, though.  Is there any way to tell Gnus to find my thread
> there?

See gnus-refer-article-method.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: chasing mail threads

[prad]

Dave Abrahams <dave@boostpro.com> writes:

> I also very commonly need to get back to look at the history of a thread.
> Gnus has `A T' and `^' which I use all the time...
>
> but it doesn't work so well in my Inbox because the messages have
> usually been deleted.
>
> I have an Archive folder that keeps all the mail I've ever sent or
> received, though.  Is there any way to tell Gnus to find my thread
> there?
>
i keep all my emails and replies in one box and recover entire threads
with
A R

-- 
in friendship,
prad

Re: chasing mail threads

[Dave Abrahams]

At Tue, 14 Jun 2011 22:57:13 +0200,
Andreas Schwab wrote:
> 
> Dave Abrahams <dave@boostpro.com> writes:
> 
> > I have an Archive folder that keeps all the mail I've ever sent or
> > received, though.  Is there any way to tell Gnus to find my thread
> > there?
> 
> See gnus-refer-article-method.

Okay, I see it.  I do appreciate the hint, but it's practically
undocumented, and I'm not familiar enough with Gnus internals to have
any idea how to take advantage of it.  Could you perhaps elaborate a
bit?

Thanks,

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Bug with referring articles (was: chasing mail threads)

[Tassilo Horn]

Dave Abrahams <dave@boostpro.com> writes:

Hi Dave,

>> See gnus-refer-article-method.
>
> Okay, I see it.  I do appreciate the hint, but it's practically
> undocumented, and I'm not familiar enough with Gnus internals to have
> any idea how to take advantage of it.  Could you perhaps elaborate a
> bit?

,----[ C-h v gnus-refer-article-method RET ]
| gnus-refer-article-method is a variable defined in `gnus.el'.
| Its value is (current
|  (nnregistry)
|  (nnweb "gmane"
| 	(nnweb-type gmane))
|  (nnweb "google"
| 	(nnweb-type google)))
| 
| Original value was current
`----

So my setting for referring articles means,

  - try to find the message in the current select method,
  - if that doesn't work, query the registry,
  - if that doesn't work, ask gmane via the web,
  - if all fails, try google groups.

The second entry requires the gnus registry, of course.  So you have to
initialize it in your ~/.gnus.el with

  (gnus-registry-initialize)

I'm not completely sure, but I have the gut feeling that `current'
should actually work for IMAP (you are using IMAP, right?).  But that's
the default value, and you say it doesn't work for you...

Ok, so now I've edebugged that stuff, and it seems there is a bug.
First, I moved some parent article from my IMAP ding group to my normal
INBOX.  Mork concretely, I've moved lee's article

  "problems with archive" / <87d3ig2qcf.fsf@yun.yagibdah.de>

from nnimap+Fastmail:INBOX.mailinglists.ding to nnimap+Fastmail:INBOX,
so that lee's

  "Re: problems with archive" / <8739jc13hi.fsf@yun.yagibdah.de>

is left alone in the ding group and hitting `^' on it will trigger
article referring.

In `gnus-summary-refer-article' there's this loop

,----[ ~/repos/el/gnus/lisp/gnus-sum.el ]
| 	(catch 'found
| 	  (dolist (gnus-override-method (gnus-refer-article-methods))
| 	    (when (and (gnus-check-server gnus-override-method)
| 		       ;; Fetch the header,
| 		       (setq number (gnus-summary-insert-subject message-id)))
| 	      ;; and display the article.
| 	      (gnus-summary-select-article nil nil nil number)
| 	      (throw 'found t)))
`----

which should try all article refer methods until one finds the referred
article.  But that doesn't work at all.

The problem is that `gnus-summary-insert-subject' doesn't find the
referred article (1), although its on the same server but a different
IMAP folder.  But instead it inserts a pseudo-article and returns 0,
thus the `when' succeeds and 'found is thrown and no other refer methods
are tried.

IMHO, the pseudo-article should be inserted only after all refer methods
failed...

(1) Is it correct that IMAP has no way to find a message by Message-ID
    in all folders?  In my case, I've just moved the referred article
    from my ding group to the INBOX on the same IMAP server and had
    expected the `current' gnus-refer-article-method to work by
    searching for that Message-ID on the server...

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles (was: chasing mail threads)

[Dave Abrahams]

Tassilo,

Thanks very much for your thorough reply; if we can solve this problem
it will make it possible for me to switch back to Gnus, which I would
dearly like to do.

At Wed, 15 Jun 2011 09:42:39 +0200,
Tassilo Horn wrote:
> 
> Dave Abrahams <dave@boostpro.com> writes:
> 
> Hi Dave,
> 
> >> See gnus-refer-article-method.
> >
> > Okay, I see it.  I do appreciate the hint, but it's practically
> > undocumented, and I'm not familiar enough with Gnus internals to have
> > any idea how to take advantage of it.  Could you perhaps elaborate a
> > bit?
> 
> ,----[ C-h v gnus-refer-article-method RET ]
> | gnus-refer-article-method is a variable defined in `gnus.el'.
> | Its value is (current
> |  (nnregistry)
> |  (nnweb "gmane"
> | 	(nnweb-type gmane))
> |  (nnweb "google"
> | 	(nnweb-type google)))
> | 
> | Original value was current
> `----
> 
> So my setting for referring articles means,
> 
>   - try to find the message in the current select method,
>   - if that doesn't work, query the registry,
>   - if that doesn't work, ask gmane via the web,
>   - if all fails, try google groups.

Ah, very nice!

> The second entry requires the gnus registry, of course.  So you have to
> initialize it in your ~/.gnus.el with
> 
>   (gnus-registry-initialize)

Sure.

> I'm not completely sure, but I have the gut feeling that `current'
> should actually work for IMAP (you are using IMAP, right?).  

Yes, I am.

> But that's the default value, and you say it doesn't work for you...

Well, it never has in the past.  I can try again...

Nope, it doesn't work.  `A T' doesn't do anything interesting, and `^'
now inserts a pseudo-article (which is a new behavior since last time
I tried this, which was a while ago).

<snip good analysis>

> The problem is that `gnus-summary-insert-subject' doesn't find the
> referred article (1), although its on the same server but a different
> IMAP folder.  But instead it inserts a pseudo-article and returns 0,
> thus the `when' succeeds and 'found is thrown and no other refer methods
> are tried.
> 
> IMHO, the pseudo-article should be inserted only after all refer methods
> failed...
> 
> (1) Is it correct that IMAP has no way to find a message by Message-ID
>     in all folders?  

I think maybe that is correct.  In my case, I'm even willing to tell
it what folder to search in, but it looks like Gnus doesn't give me a
way to express that.  I tried using the nnimap-inbox option, but that
didn't seem to lead anywhere.

Thanks again

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: chasing mail threads

[Steinar Bang]

>>>>> Dave Abrahams <dave@boostpro.com>:

> I have an Archive folder that keeps all the mail I've ever sent or
> received, though.  Is there any way to tell Gnus to find my thread
> there?

FWIW what I do is to expire to an archive folder, and I usually access
the articles in there using search with `G G' (the group is too large to
enter in a meaningful way).  Once `G G' finds something meaningful, I
use `A T' to warp to the actual archive group, and then I tick mark or
respond to the warped articles.

Re: Bug with referring articles (was: chasing mail threads)

[Dave Abrahams]

At Wed, 15 Jun 2011 04:01:22 -0400,
Dave Abrahams wrote:
> 
> > The problem is that `gnus-summary-insert-subject' doesn't find the
> > referred article (1), although its on the same server but a different
> > IMAP folder.  But instead it inserts a pseudo-article and returns 0,
> > thus the `when' succeeds and 'found is thrown and no other refer methods
> > are tried.
> > 
> > IMHO, the pseudo-article should be inserted only after all refer methods
> > failed...
> > 
> > (1) Is it correct that IMAP has no way to find a message by Message-ID
> >     in all folders?  
> 
> I think maybe that is correct.  In my case, I'm even willing to tell
> it what folder to search in, but it looks like Gnus doesn't give me a
> way to express that.  I tried using the nnimap-inbox option, but that
> didn't seem to lead anywhere.

Well, I did some hacking around and some edebugging based on what you
pointed me at, and I was able to use some advice to get
`nnimap-find-article-by-message-id' to always look in my archive
group, and it found the article, but then of course the rest of
nnimap-request-article failed.

I wonder if it would be reasonably easy to hack up a "derived" select
method based on nnimap that always forces itself to use one particular
group (mailbox)?  Then I could put that in my
gnus-refer-article-method and it would fall back to using it when the
article wasn't in my current mailbox.


-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Tassilo Horn]

Dave Abrahams <dave@boostpro.com> writes:

Hi Dave,

> Well, I did some hacking around and some edebugging based on what you
> pointed me at, and I was able to use some advice to get
> `nnimap-find-article-by-message-id' to always look in my archive
> group, and it found the article, but then of course the rest of
> nnimap-request-article failed.
>
> I wonder if it would be reasonably easy to hack up a "derived" select
> method based on nnimap that always forces itself to use one particular
> group (mailbox)?

That matches your usecase, but it's not very general...

Out of couriosity, I just process-marked all ~12 folders of one of my
imap servers in *Group* and did

  C-u G G HEADER Message-ID <87d3ig2qcf.fsf@yun.yagibdah.de> RET imap RET

to search for lee's moved message in all folders containing about 50.000
messages in total using a plain IMAP seach.  The nnir result buffer
contained this single message and the search over all groups took about
3 seconds.  So basically, I think it would be reasonable try nnir imap
searches in the `current' refer method if that is imap.  And of course,
that should be shortcutting.  In my search, actually the first group
(INBOX) contained the message I was looking for, so there was no reason
to search the other groups as well.

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Dave Abrahams]

At Wed, 15 Jun 2011 20:29:43 +0200,
Tassilo Horn wrote:
> 
> Dave Abrahams <dave@boostpro.com> writes:
> 
> Hi Dave,
> 
> > Well, I did some hacking around and some edebugging based on what you
> > pointed me at, and I was able to use some advice to get
> > `nnimap-find-article-by-message-id' to always look in my archive
> > group, and it found the article, but then of course the rest of
> > nnimap-request-article failed.
> >
> > I wonder if it would be reasonably easy to hack up a "derived" select
> > method based on nnimap that always forces itself to use one particular
> > group (mailbox)?
> 
> That matches your usecase, but it's not very general...

No it's not very general, but maybe you don't appreciate just how long
I've been looking for this capbility :-)

> Out of couriosity, I just process-marked all ~12 folders of one of my
> imap servers in *Group* and did
> 
>   C-u G G HEADER Message-ID <87d3ig2qcf.fsf@yun.yagibdah.de> RET imap RET
> 
> to search for lee's moved message in all folders containing about 50.000
> messages in total using a plain IMAP seach.  The nnir result buffer
> contained this single message and the search over all groups took about
> 3 seconds.  So basically, I think it would be reasonable try nnir imap
> searches in the `current' refer method if that is imap.  

I'm not quite sure what you're suggesting here.  A modification to Gnus, or...?

> And of course, that should be shortcutting.  In my search, actually
> the first group (INBOX) contained the message I was looking for, so
> there was no reason to search the other groups as well.

Hmm, I don't think I really looked into nnir as thoroughly as I should
have.  I have a keystroke in Wanderlust that produces the equivalent
of an nnir group containing the thread in question.  I suppose I could
do the same for Gnus.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Dave Abrahams]

At Wed, 15 Jun 2011 18:16:13 -0400,
Dave Abrahams wrote:
> 
> At Wed, 15 Jun 2011 20:29:43 +0200,
> Tassilo Horn wrote:
> >   C-u G G HEADER Message-ID <87d3ig2qcf.fsf@yun.yagibdah.de> RET imap RET
>
> Hmm, I don't think I really looked into nnir as thoroughly as I should
> have.  

It occurs to me that part of the reason is that the syntax is still a
bit of a mystery to me.  Is that documented anywhere?  I see some
vague stuff about boolean queries in the Info documentation but
nothing that mentions, e.g., `HEADER'

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Vegard Vesterheim]

On Wed, 15 Jun 2011 18:20:46 -0400 Dave Abrahams <dave@boostpro.com> wrote:

> It occurs to me that part of the reason is that the syntax is still a
> bit of a mystery to me.  Is that documented anywhere?  I see some
> vague stuff about boolean queries in the Info documentation but
> nothing that mentions, e.g., `HEADER'

I do this:
      (setq nnir-imap-default-search-key "Imap")

This means it is possible to use raw imap queries, (syntax for the
IMAP search command is described in the IMAP rfc). Then I can issue
queries like

     "from lars"
     "from lars since 20-Sep-2010"
     "from lars since 20-Sep-2010 subject imap"
     "(or from lars from vegard) since 20-Sep-2010 subject imap"
     "(not (or from lars from vegard)) since 20-Sep-2010 subject imap"
     "subject imap body query"
     "header user-agent wanderlust"

 - Vegard V -

Re: Bug with referring articles

[Tassilo Horn]

Dave Abrahams <dave@boostpro.com> writes:

>> Out of couriosity, I just process-marked all ~12 folders of one of my
>> imap servers in *Group* and did
>> 
>>   C-u G G HEADER Message-ID <87d3ig2qcf.fsf@yun.yagibdah.de> RET imap RET
>> 
>> to search for lee's moved message in all folders containing about
>> 50.000 messages in total using a plain IMAP seach.  The nnir result
>> buffer contained this single message and the search over all groups
>> took about 3 seconds.  So basically, I think it would be reasonable
>> try nnir imap searches in the `current' refer method if that is imap.
>
> I'm not quite sure what you're suggesting here.  A modification to
> Gnus, or...?

Yes, exactly.  Basically, there are two points.  Firstly, the current
code for referring articles is broken, because only the first method in
`gnus-refer-article-method' is actually tried as I've explained in
<87fwnbr15c.fsf@fastmail.fm>.  So that has to be fixed anyway.

Secondly, I think that the `current' method may perform nnir searches
over all groups of that server if the article cannot be found in the
current group and nnir is configured properly, as it is by default for
IMAP.  Or alternatively, there might be some new `current-nnir' method.

However, I have to admit that this is only some conceptual talk.  I have
no real clue how the code is supposed to work, and this whole summary
updating doesn't look too trivial. :-)

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Vegard" == Vegard Vesterheim <vegard.vesterheim@uninett.no> writes:

    Vegard> On Wed, 15 Jun 2011 18:20:46 -0400 Dave Abrahams <dave@boostpro.com> wrote:
    >> It occurs to me that part of the reason is that the syntax is
    >> still a bit of a mystery to me.  Is that documented anywhere?  I
    >> see some vague stuff about boolean queries in the Info
    >> documentation but nothing that mentions, e.g., `HEADER'

This should be documented in the gnus manual. You can search on any
email header by using ctrl-u G G which first prompts for the query
string and then the header to search in. So to search for a Message-ID
try

  C-u G G  <87d3ig2qcf.fsf@yun.yagibdah.de> RET Message-ID  RET

Alternatively if you know the imap search syntax you can use a raw imap
search (this is what Tassilo originally  suggested)

  C-u G G HEADER Message-ID <87d3ig2qcf.fsf@yun.yagibdah.de> RET imap RET

If you want to make raw imap searching the default you can follow
Vegard's suggestion

    Vegard> I do this: (setq nnir-imap-default-search-key "Imap")

although I think that as of a few months ago "Imap" should be lower all
case.

Finally note that you should be able to search all the groups under a
single topic executing the search with the the cursor on the topic line
This would avoid having to mark a bunch of groups.

Andy

Re: Bug with referring articles

[Dave Abrahams]

At Thu, 16 Jun 2011 08:01:25 +0200,
Vegard Vesterheim wrote:
> 
> On Wed, 15 Jun 2011 18:20:46 -0400 Dave Abrahams <dave@boostpro.com> wrote:
> 
> > It occurs to me that part of the reason is that the syntax is still a
> > bit of a mystery to me.  Is that documented anywhere?  I see some
> > vague stuff about boolean queries in the Info documentation but
> > nothing that mentions, e.g., `HEADER'
> 
> I do this:
>       (setq nnir-imap-default-search-key "Imap")
> 
> This means it is possible to use raw imap queries, (syntax for the
> IMAP search command is described in the IMAP rfc). Then I can issue
> queries like
> 
>      "from lars"
>      "from lars since 20-Sep-2010"
>      "from lars since 20-Sep-2010 subject imap"
>      "(or from lars from vegard) since 20-Sep-2010 subject imap"
>      "(not (or from lars from vegard)) since 20-Sep-2010 subject imap"
>      "subject imap body query"
>      "header user-agent wanderlust"

Great, thanks!  I'll try it!

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Dave Abrahams]

At Thu, 16 Jun 2011 08:51:49 +0200,
Tassilo Horn wrote:
> 
> Firstly, the current code for referring articles is broken, because
> only the first method in `gnus-refer-article-method' is actually
> tried as I've explained in <87fwnbr15c.fsf@fastmail.fm>.  So that
> has to be fixed anyway.
> 
> Secondly, I think that the `current' method may perform nnir searches
> over all groups of that server if the article cannot be found in the
> current group and nnir is configured properly, as it is by default for
> IMAP.  Or alternatively, there might be some new `current-nnir' method.
> 
> However, I have to admit that this is only some conceptual talk.  I have
> no real clue how the code is supposed to work, and this whole summary
> updating doesn't look too trivial. :-)

Is there a place where we can put this information so that it will be
treated as a bug report, and someone who has the missing clue might
respond?

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Tassilo Horn]

Dave Abrahams <dave@boostpro.com> writes:

Hi!

>> Firstly, the current code for referring articles is broken, because
>> only the first method in `gnus-refer-article-method' is actually
>> tried as I've explained in <87fwnbr15c.fsf@fastmail.fm>.  So that
>> has to be fixed anyway.
>> 
>> Secondly, I think that the `current' method may perform nnir searches
>> over all groups of that server if the article cannot be found in the
>> current group and nnir is configured properly, as it is by default for
>> IMAP.  Or alternatively, there might be some new `current-nnir' method.
>> 
>> However, I have to admit that this is only some conceptual talk.  I have
>> no real clue how the code is supposed to work, and this whole summary
>> updating doesn't look too trivial. :-)
>
> Is there a place where we can put this information so that it will be
> treated as a bug report, and someone who has the missing clue might
> respond?

I'm optimistic that Lars sends a "Done!" message the next few days
(*wink*, *wink*).  If not, I'll submit a report to the emacs bug
tracker.  (Or should one still use M-x gnus-bug?  I think, some weeks
ago Ted asked on emacs-devel if gnus bugs could be filed at emacs
debbugs as well, and that his suggestion has been accepted...)

Bye,
Tassilo

Re: Bug with referring articles

[Ted Zlatanov]

On Thu, 16 Jun 2011 21:51:24 +0200 Tassilo Horn <tassilo@member.fsf.org> wrote: 

TH> If not, I'll submit a report to the emacs bug tracker.  (Or should
TH> one still use M-x gnus-bug?  I think, some weeks ago Ted asked on
TH> emacs-devel if gnus bugs could be filed at emacs debbugs as well,
TH> and that his suggestion has been accepted...)

It's on my TODO list.  Sorry about the delay implementing this, I've
been too busy.

Ted

Re: Bug with referring articles

[Tassilo Horn]

Ted Zlatanov <tzz@lifelogs.com> writes:

Hi Ted,

> TH> If not, I'll submit a report to the emacs bug tracker.  (Or should
> TH> one still use M-x gnus-bug?  I think, some weeks ago Ted asked on
> TH> emacs-devel if gnus bugs could be filed at emacs debbugs as well,
> TH> and that his suggestion has been accepted...)
>
> It's on my TODO list.  Sorry about the delay implementing this, I've
> been too busy.

Don't worry, but tell us what's the right tool for Gnus bugs at the
moment.  M-x report-emacs-bug or gnus-bug?

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Andrew Cohen]

Just an FYI: as I recall the function gnus-summary-refer-thread was
originally written only for articles in an nntp backend. I modified it
heavily some months ago to work in nnimap groups but never got around to
implementing other refer methods. Its been on my todo list for ages. But
Ted might get to it first.

Best,
Andy

Re: Bug with referring articles

[Andrew Cohen]

I've tracked this down and I'll push a fix shortly.

andy

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

Hi Andy,

> I've tracked this down and I'll push a fix shortly.

I've tested your patch and it seems to work fine.  The `current' select
method (nnimap) fails to find the article in the current group, and
instead of inserting a pseudo article and stopping, now the second best
`gnus-refer-article-method' (nnregistry) gets a chance and finds the
article in my INBOX.

So the bug is fixed.  Thanks a lot! :-)

What remains as a feature suggestion is that the nnimap backend could
check the other (subscribed) IMAP folders of that select method, too, if
the current folder doesn't contain the article one tries to refer.

The benefit over the gnus registry is that it'll work even for articles
you didn't read with Gnus (and thus are not registered) and with
articles you've moved on the server using some other mail client, like
your mail provider's web frontend or the mail app of your smartphone.

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Lars Magne Ingebrigtsen]

Tassilo Horn <tassilo@member.fsf.org> writes:

> What remains as a feature suggestion is that the nnimap backend could
> check the other (subscribed) IMAP folders of that select method, too, if
> the current folder doesn't contain the article one tries to refer.

Yeah, if nnir is able to find the article, then there should be a way to
make `^' hook up to that.

Looking at the code, wouldn't just changing this to do a search:

(deffoo nnir-request-article (article &optional group server to-buffer)
  (if (stringp article)
      (nnheader-report
       'nnir
       "nnir-retrieve-headers doesn't grok message ids: %s"
       article)
  ...))

and then putting the nnir select method into
`gnus-refer-article-method' work?  Er, no, we'd probably have to have a
special `current-nnir' thing like you said...

It would also be nice to search the archive backend (by default), but
since that's nnfolder, which doesn't really support that, then that
kinda sucks...

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

Re: Bug with referring articles

[Andrew Cohen]

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

    Lars> Tassilo Horn <tassilo@member.fsf.org> writes:
    >> What remains as a feature suggestion is that the nnimap backend
    >> could check the other (subscribed) IMAP folders of that select
    >> method, too, if the current folder doesn't contain the article
    >> one tries to refer.

    Lars> Yeah, if nnir is able to find the article, then there should
    Lars> be a way to make `^' hook up to that.

Yup, I've been playing with it.

[...]


    Lars> and then putting the nnir select method into
    Lars> `gnus-refer-article-method' work?  Er, no, we'd probably have
    Lars> to have a special `current-nnir' thing like you said...

Yeah, this is a problem. A very simple (but less automatic :()
alternative would be to specify the server directly : (nnir
"nnimap:name"). Then you could even search for a referring article that
is on a different server (which happens to me occasionally). I've hacked
this up if anyone wants to try it out.

Regards,
Andy

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

Hi Andy,

>     Lars> and then putting the nnir select method into
>     Lars> `gnus-refer-article-method' work?  Er, no, we'd probably have
>     Lars> to have a special `current-nnir' thing like you said...
>
> Yeah, this is a problem. A very simple (but less automatic :()
> alternative would be to specify the server directly : (nnir
> "nnimap:name"). Then you could even search for a referring article
> that is on a different server (which happens to me occasionally). I've
> hacked this up if anyone wants to try it out.

Sure thing, I'm happy to give it a try.  But probably I will only find
some time to do that on the weekend.

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Ted Zlatanov]

On Fri, 17 Jun 2011 09:30:57 +0200 Tassilo Horn <tassilo@member.fsf.org> wrote: 

TH> Ted Zlatanov <tzz@lifelogs.com> writes:
TH> Hi Ted,

TH> If not, I'll submit a report to the emacs bug tracker.  (Or should
TH> one still use M-x gnus-bug?  I think, some weeks ago Ted asked on
TH> emacs-devel if gnus bugs could be filed at emacs debbugs as well,
TH> and that his suggestion has been accepted...)
>> 
>> It's on my TODO list.  Sorry about the delay implementing this, I've
>> been too busy.

TH> Don't worry, but tell us what's the right tool for Gnus bugs at the
TH> moment.  M-x report-emacs-bug or gnus-bug?

Either will work, but `report-emacs-bug' is probably better since we're
moving that way anyhow.

Ted

Re: Bug with referring articles

[Tassilo Horn]

Ted Zlatanov <tzz@lifelogs.com> writes:

Hi Ted,

>>> It's on my TODO list.  Sorry about the delay implementing this, I've
>>> been too busy.
>
> TH> Don't worry, but tell us what's the right tool for Gnus bugs at
> TH> the moment.  M-x report-emacs-bug or gnus-bug?
>
> Either will work, but `report-emacs-bug' is probably better since
> we're moving that way anyhow.

Good to know, but Andy already fixed the bug I wanted to report. ;-)

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Ted Zlatanov]

On Tue, 28 Jun 2011 20:01:15 +0200 Tassilo Horn <tassilo@member.fsf.org> wrote: 

TH> Good to know, but Andy already fixed the bug I wanted to report. ;-)

You guys are so impatient, can't wait 3 months! :)

Ted

Re: Bug with referring articles

[Lars Magne Ingebrigtsen]

Andrew Cohen <cohen@andy.bu.edu> writes:

>     Lars> and then putting the nnir select method into
>     Lars> `gnus-refer-article-method' work?  Er, no, we'd probably have
>     Lars> to have a special `current-nnir' thing like you said...
>
> Yeah, this is a problem. A very simple (but less automatic :()
> alternative would be to specify the server directly : (nnir
> "nnimap:name"). Then you could even search for a referring article that
> is on a different server (which happens to me occasionally). I've hacked
> this up if anyone wants to try it out.

Please apply, so that people can test it out.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

Re: Bug with referring articles

[Andrew Cohen]

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

    Lars> Please apply, so that people can test it out.  :-)


OK, if you insist:)

Comments:

1. You can add nnir to the list of referral methods as (nnir
"server"). If the server part is absent (e.g. (nnir)) then the current server
is used. My value for gnus-refer-article-method is

gnus-refer-article-method is a variable defined in `gnus.el'.
Its value is (current
 (nnregistry)
 (nnir)
 (nnir "gmail"))


2. Imap searching doesn't span groups, so nnir just searches each group
on the server in turn. This might be slow (although its pretty fast for
me).

3. The function `gnus-summary-refer-article' first tries to find the
referent article's header and insert a line in the summary buffer
(gnus-summary-insert-subject) and then selects the article
(gnus-summary-select-article). This has the effect of looping through
the referral method twice. I've memoized things so this doesn't happen
for the nnir referral, but with some more thought the double looping can
probably be avoided all-together. 

Re: Bug with referring articles

[Dave Abrahams]

At Thu, 16 Jun 2011 21:51:24 +0200,
Tassilo Horn wrote:
> 
> Dave Abrahams <dave@boostpro.com> writes:
> 
> Hi!
> 
> >> Firstly, the current code for referring articles is broken, because
> >> only the first method in `gnus-refer-article-method' is actually
> >> tried as I've explained in <87fwnbr15c.fsf@fastmail.fm>.  So that
> >> has to be fixed anyway.
> >> 
> >> Secondly, I think that the `current' method may perform nnir searches
> >> over all groups of that server if the article cannot be found in the
> >> current group and nnir is configured properly, as it is by default for
> >> IMAP.  Or alternatively, there might be some new `current-nnir' method.
> >> 
> >> However, I have to admit that this is only some conceptual talk.  I have
> >> no real clue how the code is supposed to work, and this whole summary
> >> updating doesn't look too trivial. :-)
> >
> > Is there a place where we can put this information so that it will be
> > treated as a bug report, and someone who has the missing clue might
> > respond?
> 
> I'm optimistic that Lars sends a "Done!" message the next few days
> (*wink*, *wink*).  If not, I'll submit a report to the emacs bug
> tracker.  (Or should one still use M-x gnus-bug?  I think, some weeks
> ago Ted asked on emacs-devel if gnus bugs could be filed at emacs
> debbugs as well, and that his suggestion has been accepted...)

Bump.  Let's not lose track of this!

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Lars Magne Ingebrigtsen]

Dave Abrahams <dave@boostpro.com> writes:

> Bump.  Let's not lose track of this!

Andrew checked in code for doing this earlier today, and would probably
be happy to get testers.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Dave" == Dave Abrahams <dave@boostpro.com> writes:

    >> >
    >> > Is there a place where we can put this information so that it
    >> will be > treated as a bug report, and someone who has the
    >> missing clue might > respond?
    >> 
    >> I'm optimistic that Lars sends a "Done!" message the next few
    >> days (*wink*, *wink*).  If not, I'll submit a report to the emacs
    >> bug tracker.  (Or should one still use M-x gnus-bug?  I think,
    >> some weeks ago Ted asked on emacs-devel if gnus bugs could be
    >> filed at emacs debbugs as well, and that his suggestion has been
    >> accepted...)

    Dave> Bump.  Let's not lose track of this!


Check out my posts in the rest of this thread. I checked in something
for this a while ago, and am just waiting for some test results.

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

Hi Andrew,

>     Dave> Bump.  Let's not lose track of this!
>
> Check out my posts in the rest of this thread. I checked in something
> for this a while ago, and am just waiting for some test results.

I've moved some messages via the web frontend of my provider, so that
the registry doesn't get find them any more.  Your nnir method searched
all groups of that server and fetched the correct message.  Awesome!

Two minor things: it seems the search is not short-cutting, and the
docstring of `gnus-refer-article-method' doesn't mention nnir, yet.

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Tassilo" == Tassilo Horn <tassilo@member.fsf.org> writes:

    Tassilo> Andrew Cohen <cohen@andy.bu.edu> writes: Hi Andrew,


    Tassilo> I've moved some messages via the web frontend of my
    Tassilo> provider, so that the registry doesn't get find them any
    Tassilo> more.  Your nnir method searched all groups of that server
    Tassilo> and fetched the correct message.  Awesome!

Great!

    Tassilo> Two minor things: it seems the search is not short-cutting,

Do you mean that the nnir search is performed twice in a row? Or do you
mean something else?

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

Hi Andy,

>     Tassilo> Two minor things: it seems the search is not short-cutting,
>
> Do you mean that the nnir search is performed twice in a row? Or do
> you mean something else?

No, I've meant that it searches all groups and doesn't stop when the
message was found.  For example, in the search below the message was
correcly found in the bbdb-info group, but it searches all following
groups anyway.

--8<---------------cut here---------------start------------->8---
nnregistry: requesting article `<87d3ig2qcf.fsf@yun.yagibdah.de>' in group `nnimap+Fastmail:INBOX'
20110702T000832.537> Opening nnir server on nnimap:Fastmail...
20110702T000832.538> Opening nnir server on nnimap:Fastmail...done
Opening server Fastmail
Searching nnimap+Fastmail:INBOX.uni-old.diplomarbeit...done
Searching nnimap+Fastmail:INBOX.uni-old...done
Searching nnimap+Fastmail:INBOX.training.spam...done
Searching nnimap+Fastmail:INBOX.training.ham...done
Searching nnimap+Fastmail:INBOX.training...done
Searching nnimap+Fastmail:INBOX.mailinglists.misc...done
Searching nnimap+Fastmail:INBOX.mailinglists.info-fsf...done
Searching nnimap+Fastmail:INBOX.mailinglists.golang-nuts...done
Searching nnimap+Fastmail:INBOX.mailinglists.gnome-shell...done
Searching nnimap+Fastmail:INBOX.mailinglists.emacs-devel...done
Searching nnimap+Fastmail:INBOX.mailinglists.ding...done
Searching nnimap+Fastmail:INBOX.mailinglists.d-barf...done
Searching nnimap+Fastmail:INBOX.mailinglists.clojure-de...done
Searching nnimap+Fastmail:INBOX.mailinglists.clojure...done
Searching nnimap+Fastmail:INBOX.mailinglists.bbdb-info...done
Searching nnimap+Fastmail:INBOX.mailinglists...done
Searching nnimap+Fastmail:INBOX.bugs...done
Searching nnimap+Fastmail:INBOX.Trash...done
Searching nnimap+Fastmail:INBOX.Sent Items...done
Searching nnimap+Fastmail:INBOX.Junk Mail...done
Searching nnimap+Fastmail:INBOX.FSF...done
Searching nnimap+Fastmail:INBOX.Drafts...done
Searching nnimap+Fastmail:INBOX.Chats...done
Searching nnimap+Fastmail:INBOX...done
--8<---------------cut here---------------end--------------->8---

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Tassilo" == Tassilo Horn <tassilo@member.fsf.org> writes:

    >> 
    >> Do you mean that the nnir search is performed twice in a row? Or
    >> do you mean something else?

    Tassilo> No, I've meant that it searches all groups and doesn't stop
    Tassilo> when the message was found.  

Err, yeah, that's the "expected" behavior. All the search functions are
designed this way. I can probably modify them to take a shortcut, but
they don't at the moment. It shouldn't be that hard to do, but is it
worth it? My searches (admittedly with only a few dozen groups) finish
quickly enough that I don't think I would notice the difference.

Andy

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

>     Tassilo> No, I've meant that it searches all groups and doesn't stop
>     Tassilo> when the message was found.  
>
> Err, yeah, that's the "expected" behavior. All the search functions
> are designed this way. I can probably modify them to take a shortcut,
> but they don't at the moment. It shouldn't be that hard to do, but is
> it worth it? My searches (admittedly with only a few dozen groups)
> finish quickly enough that I don't think I would notice the
> difference.

Yes, it's quite fast to find one article (say, 2 or 3 secs), but
short-cutting might be worth when having to search for many different
articles during `A T'.  For example, some people have folders like
2011/jan, 2011/feb, so referring threads wrapping around months is
costly.  (Not that I say that's a good way of organizing mail...)

Well, I'm happy with how it works right now.  I've just thought the
non-shortcutting was not indended.

Bye,
Tassilo

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Tassilo" == Tassilo Horn <tassilo@member.fsf.org> writes:

    >> 
    >> Err, yeah, that's the "expected" behavior. All the search
    >> functions are designed this way. I can probably modify them to
    >> take a shortcut, but they don't at the moment. It shouldn't be
    >> that hard to do, but is it worth it? My searches (admittedly with
    >> only a few dozen groups) finish quickly enough that I don't think
    >> I would notice the difference.

    Tassilo> Yes, it's quite fast to find one article (say, 2 or 3
    Tassilo> secs), but short-cutting might be worth when having to
    Tassilo> search for many different articles during `A T'.  For
    Tassilo> example, some people have folders like 2011/jan, 2011/feb,
    Tassilo> so referring threads wrapping around months is costly.
    Tassilo> (Not that I say that's a good way of organizing mail...)

Since it's easy enough to do, I went ahead and did it. Give it a try and
let me know if it works. 

By the way, if searching one group is taking a long time you /should/ be
able to C-g to move on to the next group in the list.


Andy

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Tassilo" == Tassilo Horn <tassilo@member.fsf.org> writes:


    Tassilo> and the docstring of `gnus-refer-article-method' doesn't
    Tassilo> mention nnir, yet.

The docstring of `gnus-refer-article-method' begins:

*Preferred method for fetching an article by Message-ID.
If you are reading news from the local spool (with nnspool), fetching
articles by Message-ID is painfully slow.  By setting this method to an
nntp method, you might get acceptable results.*

This seems a tad outdated? Maybe this is no longer the most important
use-case for this variable, and shouldn't be the first thing the user is
confronted with?

Andy

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

Hi Andy,

> The docstring of `gnus-refer-article-method' begins:
>
> *Preferred method for fetching an article by Message-ID.
> If you are reading news from the local spool (with nnspool), fetching
> articles by Message-ID is painfully slow.  By setting this method to an
> nntp method, you might get acceptable results.*
>
> This seems a tad outdated? Maybe this is no longer the most important
> use-case for this variable, and shouldn't be the first thing the user
> is confronted with?

I'd agree.  I guess the most relevant backends are imap, nntp, and
probably nnml.

Bye,
Tassilo
-- 
Sent from my Emacs

Re: Bug with referring articles

[Tassilo Horn]

Andrew Cohen <cohen@andy.bu.edu> writes:

>     Tassilo> Yes, it's quite fast to find one article (say, 2 or 3
>     Tassilo> secs), but short-cutting might be worth when having to
>     Tassilo> search for many different articles during `A T'.  For
>     Tassilo> example, some people have folders like 2011/jan, 2011/feb,
>     Tassilo> so referring threads wrapping around months is costly.
>     Tassilo> (Not that I say that's a good way of organizing mail...)
>
> Since it's easy enough to do, I went ahead and did it. Give it a try
> and let me know if it works.

Great, it works like a charm. :-)

> By the way, if searching one group is taking a long time you /should/
> be able to C-g to move on to the next group in the list.

It's all to fast to test that. ;-)

Bye,
Tassilo

Re: Bug with referring articles

[Dave Abrahams]

on Sun Jun 26 2011, Lars Magne Ingebrigtsen <larsi-AT-gnus.org> wrote:

> Tassilo Horn <tassilo@member.fsf.org> writes:
>
>> What remains as a feature suggestion is that the nnimap backend could
>> check the other (subscribed) IMAP folders of that select method, too, if
>> the current folder doesn't contain the article one tries to refer.
>
> Yeah, if nnir is able to find the article, then there should be a way to
> make `^' hook up to that.
>
> Looking at the code, wouldn't just changing this to do a search:
>
> (deffoo nnir-request-article (article &optional group server to-buffer)
>   (if (stringp article)
>       (nnheader-report
>        'nnir
>        "nnir-retrieve-headers doesn't grok message ids: %s"
>        article)
>   ...))
>
> and then putting the nnir select method into
> `gnus-refer-article-method' work?  Er, no, we'd probably have to have a
> special `current-nnir' thing like you said...

Well, a big +1 for that idea, whatever it takes to implement it.

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Dave Abrahams]

on Thu Jun 30 2011, Andrew Cohen <cohen-AT-andy.bu.edu> wrote:

>>>>>> "Dave" == Dave Abrahams <dave@boostpro.com> writes:
>
>     >> >
>     >> > Is there a place where we can put this information so that it
>     >> will be > treated as a bug report, and someone who has the
>     >> missing clue might > respond?
>     >> 
>     >> I'm optimistic that Lars sends a "Done!" message the next few
>     >> days (*wink*, *wink*).  If not, I'll submit a report to the emacs
>     >> bug tracker.  (Or should one still use M-x gnus-bug?  I think,
>     >> some weeks ago Ted asked on emacs-devel if gnus bugs could be
>     >> filed at emacs debbugs as well, and that his suggestion has been
>     >> accepted...)
>
>     Dave> Bump.  Let's not lose track of this!
>
>
> Check out my posts in the rest of this thread. I checked in something
> for this a while ago, and am just waiting for some test results.

I keep finding new parts of this thread that I haven't seen before (!)

Trying to catch up now...


-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Dave Abrahams]

on Thu Jun 30 2011, Andrew Cohen <cohen-AT-andy.bu.edu> wrote:

>>>>>> "Lars" == Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
>
>     Lars> Please apply, so that people can test it out.  :-)
>
>
> OK, if you insist:)
>
> Comments:
>
> 1. You can add nnir to the list of referral methods as (nnir
> "server"). If the server part is absent (e.g. (nnir)) then the current server
> is used. My value for gnus-refer-article-method is
>
> gnus-refer-article-method is a variable defined in `gnus.el'.
> Its value is (current
>  (nnregistry)
>  (nnir)
>  (nnir "gmail"))
>
> 2. Imap searching doesn't span groups, so nnir just searches each group
> on the server in turn. This might be slow (although its pretty fast for
> me).

This is really nice!  I can get this to work for my current server by
setting `gnus-refer-article-method' to:

  (current (nnir))

[However, I note that the customization interface doesn't give me any way
to produce that value]

The first problem with this is that it searches all the groups (of which
I happen to have many).  Since I'm using Gmail's servers, there's one
group (called [Gmail]/All Mail) where it can find every message.  Is
there a way to get it to look in just that one group?

Secondly, I couldn't figure out what should go in the element after nnir
if I wanted to specify that server directly.  Here's my
`gnus-select-method':

  (nnimap "BoostPro" (nnimap-address "imap.gmail.com") (nnimap-stream ssl) (nnimap-authenticator login) (nnimap-nov-is-evil t))

Lastly, is there any way to get this lookup to try the agent first?  If
I have the article on my local machine and I'm disconnected, it seems
like I ought to be able to get to it still.

All assistance greatly appreciated,

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Dave" == Dave Abrahams <dave@boostpro.com> writes:


    Dave> This is really nice!  I can get this to work for my current
    Dave> server by setting `gnus-refer-article-method' to:

    Dave>   (current (nnir))

    Dave> [However, I note that the customization interface doesn't give
    Dave> me any way to produce that value]

I'll get around to adding the customize stuff at some point.

    Dave> The first problem with this is that it searches all the groups
    Dave> (of which I happen to have many).  Since I'm using Gmail's
    Dave> servers, there's one group (called [Gmail]/All Mail) where it
    Dave> can find every message.  Is there a way to get it to look in
    Dave> just that one group?

Certain group scan be eliminated from nnir searching by using the
variable `nnir-ignored-newsgroups'. This is a regex so you should be
able to configure it to do what you want. However you might consider
doing the opposite (leave out the "All Mail" group and search the
rest). Although possibly slower, this will keep the information about
what group each found message is in which could be helpful if you do
much sorting of messages into different groups.

    Dave> Secondly, I couldn't figure out what should go in the element
    Dave> after nnir if I wanted to specify that server directly.
    Dave> Here's my `gnus-select-method':

Each element in `gnus-refer-article-method' is a select method (with the
exception of the special symbol `current'). You can read about select
methods in the manual. The specification of an nnir select method has
the usual form (nnir "server").  Leaving out the server simply defaults
to searching the server from which the article-referral process was
called.


    Dave>   (nnimap "BoostPro" (nnimap-address "imap.gmail.com")
    Dave> (nnimap-stream ssl) (nnimap-authenticator login)
    Dave> (nnimap-nov-is-evil t))

This is a valid select method for an nnimap server. It will have nothing
to do with searching with nnir.

    Dave> Lastly, is there any way to get this lookup to try the agent
    Dave> first?  If I have the article on my local machine and I'm
    Dave> disconnected, it seems like I ought to be able to get to it
    Dave> still.

I don't know much about the agent. I believe that the agent is is set on
a per-server basis. Since nnir is about searching only (it produces
"ephemeral" groups) it can't be agentized.

Regards,
Andy

Re: Bug with referring articles

[Dave Abrahams]

on Wed Jul 13 2011, Andrew Cohen <cohen-AT-andy.bu.edu> wrote:

>>>>>> "Dave" == Dave Abrahams <dave@boostpro.com> writes:
>
>
>     Dave> This is really nice!  I can get this to work for my current
>     Dave> server by setting `gnus-refer-article-method' to:
>
>     Dave>   (current (nnir))
>
>     Dave> [However, I note that the customization interface doesn't give
>     Dave> me any way to produce that value]
>
> I'll get around to adding the customize stuff at some point.

OK.  I think, while you're at it, some documentation is needed of what
the "Options" field means when setting up nnir there.

>     Dave> The first problem with this is that it searches all the groups
>     Dave> (of which I happen to have many).  Since I'm using Gmail's
>     Dave> servers, there's one group (called [Gmail]/All Mail) where it
>     Dave> can find every message.  Is there a way to get it to look in
>     Dave> just that one group?
>
> Certain group scan be eliminated from nnir searching by using the
> variable `nnir-ignored-newsgroups'. This is a regex so you should be
> able to configure it to do what you want. 

Ah, thanks!

> However you might consider doing the opposite (leave out the "All
> Mail" group and search the rest). Although possibly slower, this will
> keep the information about what group each found message is in which
> could be helpful if you do much sorting of messages into different
> groups.

Not useful enough to justify the slowdown.  Seriously, it's **really**
slow.  I have too many groups.  I mostly don't sort messages anymore;
most of what's there is legacy stuff.

>     Dave> Secondly, I couldn't figure out what should go in the element
>     Dave> after nnir if I wanted to specify that server directly.
>     Dave> Here's my `gnus-select-method':
>
> Each element in `gnus-refer-article-method' is a select method (with the
> exception of the special symbol `current'). You can read about select
> methods in the manual. 

I'm familiar with them, thanks.

> The specification of an nnir select method has the usual form (nnir
> "server").  Leaving out the server simply defaults to searching the
> server from which the article-referral process was called.

Yes, my problem is how to specify the "server" element.

>     Dave>   (nnimap "BoostPro" (nnimap-address "imap.gmail.com")
>     Dave> (nnimap-stream ssl) (nnimap-authenticator login)
>     Dave> (nnimap-nov-is-evil t))
>
> This is a valid select method for an nnimap server. It will have nothing
> to do with searching with nnir.

Inasmuch as it describes my server, I thought it would help you describe
what should go in the "server" element.

>     Dave> Lastly, is there any way to get this lookup to try the agent
>     Dave> first?  If I have the article on my local machine and I'm
>     Dave> disconnected, it seems like I ought to be able to get to it
>     Dave> still.
>
> I don't know much about the agent. I believe that the agent is is set on
> a per-server basis. Since nnir is about searching only (it produces
> "ephemeral" groups) it can't be agentized.

Interestingly, it does seem to change the behavior of nnir: 
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=9058

-- 
Dave Abrahams
BoostPro Computing
http://www.boostpro.com

Re: Bug with referring articles

[Andrew Cohen]

>>>>> "Dave" == Dave Abrahams <dave@boostpro.com> writes:


    Dave> Yes, my problem is how to specify the "server" element.

    Dave> (nnimap "BoostPro" (nnimap-address "imap.gmail.com")
    Dave> (nnimap-stream ssl) (nnimap-authenticator login)
    Dave> (nnimap-nov-is-evil t))
    >> 
    >> This is a valid select method for an nnimap server. It will have
    >> nothing to do with searching with nnir.

    Dave> Inasmuch as it describes my server, I thought it would help
    Dave> you describe what should go in the "server" element.

Ahh, I misunderstood. Try (nnir "nnimap:BoostPro") (in general you can
get the server corresponding to a select method using
`gnus-method-to-server'). 

    Dave> Lastly, is there any way to get this lookup to try the agent
    Dave> first?  If I have the article on my local machine and I'm
    Dave> disconnected, it seems like I ought to be able to get to it
    Dave> still.
    >> 
    >> I don't know much about the agent. I believe that the agent is is
    >> set on a per-server basis. Since nnir is about searching only (it
    >> produces "ephemeral" groups) it can't be agentized.

    Dave> Interestingly, it does seem to change the behavior of nnir:
    Dave> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=9058

Ahh, again I misunderstood. nnir searching (for imap groups) talks
directly with the imap server to find the relevant articles. Once found,
nnir passes the info off to the /regular/ nnimap select method to handle
retrieving the article. I guess this second step is somehow running
afoul of the agent. Probably I need to set up some agent variables
between the two steps. I'll look into this (but probably not soon).