Re: ognus

[Sean Doran <smd@ebone.net>]

Dave Love <d.love@dl.ac.uk> writes:

> I doubt many of them have absentee maintainers, a lack of management
> and not even the ability to make a release or update a web site.

Hm, I think it would be as easy as one of the usual
suspects making a candidate release tag and fixing up
things in that tagged branch until it seems "good enough"
for a release, tagging the release, putting tarballs up on
ftp.gnus.org, and doing the normal announcement.  If he's
around, some of all of this could be done by Lars, but I
don't think his being buried stalls most of the hard work
(i.e., everything up to the posting of the announcement &
the installation of the tarball on the ftp site).

The question is then what to do about pulling-up bugfixes
into the release; obviously fingers should get broken for
pulling-up anything incompatible into a release tag, and
to avoid unnecessary broken-fingers, someone should play
"releng", maintaining purity, making tarballs of patches
against the released tarball, and so forth.  Various
software projects do this somewhat differently; pull-ups
against the release tag and pull-ups against a second
release_bugfix tag or equivalent seem to work reasonably well.

As to the website and ftp archive, one of the usual
suspects probably could convince him to cough over write
access to his playpens, if they can do a reasonable
imitation of his whimsicalness. :-)

> The fact that this exchange is happening is indicative.  I shouldn't
> be doing the work I am and I certainly shouldn't find things change
> under me like this without warning.  (I'm not a beginner.)

Right, beginners go to the ftp archive and observe
"gnus.tar.gz", or they end up at
http://www.gnus.org/distribution.html where they see
useful information starting with, "Gnus is distributed as
a compressed tar file of Emacs Lisp files".

Perhaps the "development sources" further down the page is
insufficiently scary to people who aren't familiar with
how CVS tends to work in large multi-person projects?

> So people wouldn't advise users to do that?  

No, users should not run development sources, unless they
are prepared to deal with unexpected bugs.

> How should things be stabilized and users provided with
> useful fixed versions?  (I know stabilization prior to
> Gnus releases isn't traditional.)

"Gnus is distributed as a compressed tar file of Emacs
Lisp files".

Unfortunately, the "An archive of patches is also available"
is bit-rotting, but this would be a good place to put
pull-up fixes against the compressed tar files.

Gnus used to be released with such frequency, that
inter-release pull-ups were mostly pointless.  Releasing
with that frequency again is another option, and cuts down
on the releng activity.

>  PA> Aren't all the 5.8.7 point releases stable?
> I don't know what that means.

He means, the "gnus.tar.gz" files found in the primary and
mirror archives are stable.   The current gnus.tar.gz
contains sources to version "Gnus v5.8.7".

> Certainly things after 5.8.7 don't look too stable and 5.8.7 was
> apparently released with mail-source-delete-incoming turned off, for
> instance, even if there's notthing else seriously wrong with it.

After 5.8.7 is "Gnus development sources".

mail-source-delete-incoming -> nil is a safety feature for
new users who have never set up Gnus before, and is a
smart idea.   Anyone who has been using gnus for a while
(e.g., 5.8.6, 5.8.5, ...) may well have (setq mail-source-delete-incoming t)
in her or his .gnus.el

        Sean.