Re: CVS is the `released version'

[David Kastrup <dak@gnu.org>]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>>> The biggest benefit I see is that packages can be updated on the
>>> package maintainer's schedule.  Emacs, on the other hand, is updated
>>> on the Emacs maintainers' schedule.
>
> Pleae, let's not link the issue of including a package-installer (such as
> package.el) with the issue of which packages to (un)bundle.
> These are completely separate questions.
>
> package.el is useful to deal with non-bundled packages: there are
> many of those and they're not about to be all bundled.

I would think that if we could be able to say "if you have organized
your Lisp source files in such&such a way with this&that information
in the following format in its head, then using the command
M-x package-pull-into-current RET filename.el RET
will install the file and its associated files into your currently used
Emacs.  Using the command
M-x package-pull-into-source RET filename.el RET
will make it a part of your Emacs source tree (which you can then
either compile or commit).  And
M-x package-update RET filename.el RET
will look up any dedicated servers declared in the comments from
filename.el and will replace it (and its associated files) by a new
version which you can then either pull into current or pull into
source.

That would be a good feature/thing/whatever, and it could be made to
work with everything included in Emacs.  It could even ask whether you
want to pull from some package server, the last release from the
declared upstream, or a vc version from.

Creating a bunch of easy to follow conventions for this kind of
functionality, and updating source files to follow them by and by
would be a fine idea, I think.

But everything prescribing rigid procedures, policies, servers etc to
satisfy (like the XEmacs package system and its associated policies)
is asking for organized stagnation.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum