Re: CVS is the `released version'

[Richard Stallman]

    package.el is attached to a web site, ELPA, where package updates are
    uploaded.

In that case, I think the real proposal is not "add a package system
to Emacs" but rather "set up a standard site for Emacs add-ons".

If the add-ons are put in such a web site, finding and installing them
would be much easier.  Maybe it is worth doing that, though calling it
a "package system" seems like hype.

But there are two important non-technical problems with this approach.

1. It could reduce the incentive for people to assign copyright on
their code.

2. It would mean that Emacs refers people very strongly to a site that
isn't run by the GNU Project.  I don't know what their policies are.
But even if they are good, now, we have no way to assure that remains
so.

These problems would be eliminated if we put the package repository on
gnu.org and limit it to packages that are copyright FSF.

In other words, I can see that installing packages in a separate
repository and releasing them there, but I don't want this to
alter the legal and publicity arrangements that we would have
made for including them in Emacs.

      Packages
    are downloaded (including their dependencies, if any) and installed
    for you, autoloads are extracted, the package is byte-compiled, and
    when Emacs starts up,the packages are "activated" (meaning the
    autoloads are evalled).

It sounds convenient.  As long as it doesn't require a lot of change
in how you write the Lisp programs, I have nothing against this.

      It also mandates a
    couple file names, the main one being the .el file that holds the
    "define-package" call.

Something like a "define-package" call is one of the things that makes
me dislike package systems.  Can we avoid this?  Why is it needed?