Re: nndebbugs

[Lars Magne Ingebrigtsen <larsi@gnus.org>]

Russ Allbery <rra@stanford.edu> writes:

> Hm, I wonder if there's any way to generate an "article" for each bug,
> which if you open it is treated like a digest of the bug discussion.
> Otherwise, I think the threading between bugs and inside bugs is going to
> make matters rather confusing.

If the messages have References and stuff, it should work as a normal
mail group.  If not, it'll still gather threads based on Subject...  so
it shouldn't be more confusing than a mailing list from 1992.  :-)

> I don't know what the Emacs SOAP implementation looks like, but way faster
> than parsing the HTML output there to get a list of bugs would be the
> elisp equivalent of:
>
> #!/usr/bin/perl
> use SOAP::Lite;
> our $URL       = 'http://debbugs.gnu.org/cgi/soap.cgi';
> our $NAMESPACE = 'Debbugs/SOAP/1';
> my $soap = SOAP::Lite->uri ($NAMESPACE)->proxy ($URL);
> $soap->transport->env_proxy;
> my $bugs = $soap->get_bugs (package => 'gnus')->result;
> print join ("\n", @$bugs);
>
> That SOAP call returns an array of all the bug numbers assigned to that
> package.  You can also filter on status, of course.
>
> I just tested the above Perl code and it does the right thing with the GNU
> debbugs.

Right.  I don't think Emacs has SOAP support at all.  Could you have
your perl script dump out the XML structure, and perhaps Gnus could just
send that over "as is" to the server?

On the other hand, scraping the HTML to find the bug numbers is really
easy.  All the links are like

  http://debbugs.gnu.org/cgi/bugreport.cgi?bug=50

so it just has to get the HTML, look for "bug=[0-9]+", and there you
have it...  

> If you do this as an "article" per bug, you'd only have to fetch when
> someone actually read that article.

Gnus backends don't really support "deferred" fetching of stuff like
that...

I think one-group-per-component makes sense.  I mean, the important
thing is that you can read all new comments on all the bugs you're
interested in, isn't it?  If you have to take an extra step to inspect
each bug, it's kinda ... boring...

However, I've done some benchmarking, and if there's 100 bugs in your
set, then entering the "full" group will take 45 seconds on the Emacs
tracker.  That's way too slow to be useful.

Does the SOAP interface by any chance return more metadata -- like, for
instance, how many replies there have been to each bug, or the last time
it was updated, or anything else that can be used by nndebbugs to say
"no, I don't need to download that one, since I've already done so once
already".

Alternatively again, nndebbugs could return a list of groups -- one per
bug entry.  Then updating any single one would be fast.  But there would
be a whole lot of groups, and that's not nice, either...

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