Re: CVS is the `released version'

[David Kastrup <dak@gnu.org>]

"Robert J. Chassell" <bob@rattlesnake.com> writes:

> Remember, for many people, but not for many moderns, CVS provides the
> `released version' of GNU Emacs.  It is the prime version used by
> those who contribute.  It is not a binary that cannot be changed.
> Once in a while, development needs to be frozen while bugs are fixed.
>
> Since more and more people are coming to think of or are forced to use
> big numbered releases and eschew the daily releases, they view a delay
> between one big numbered release and another as bad.  But those who
> enjoy daily releases hardly notice.
>
> I suspect we are seeing a conflict between cultures: on the one hand,
> those who install a new release of GNU Emacs every day or almost every
> day and, on the other hand, those who look for big numbered releases,
> such as that from Emacs 21 to Emacs 22.
>     
> The world is tending towards those who look for big numbered releases
> even though most contributions are small.

I can't find I agree even remotely.  A release is a point of
stability, one where one tries to make reasonably sure that the
overall consistency (of packages working together and with the core,
and of code and documentation) is in reasonable shape, for whoever
happens to use the software.

There are development phases where most changes are incremental and
where it becomes a reasonable operation to track the trunk.  The Linux
kernel development has changed its mode of operation and its
definition of "stable" and "release" a lot over the years.  Emacs
development has pretty much stalled in release anticipation, leading
to a version where one could have made a release pretty much anytime
in the last two years and have the users get something _better_ than
they are using now, and the developers a new point from which to
continue.

Architectural changes have not happened in the last few years, even
though some architectures have had a lot of problems shaken out
(Windows and MacOSX, two proprietary platforms).  Technologies like
emacs-unicode2, and even somewhat less radical changes like multitty
are not fit to be taken in mid-stride.  While the emacs-unicode2
branch is currently comparatively stable, this is not an accident but
the result of a _lot_ of duplicate work over years because of the
pent-up release.

There are no easy answers: XEmacs has decoupled packages and releases
them separately and often.  I am not convinced of the results: there
appears to be quite some bit rot happening, the beta core has been
unstable for a large time, anyway, and the packages are often not
updated or fixed for long amounts of time.

But never releasing anything for which one has at least some
inclination to stand behind it and call it "this is as good as it gets
right now" is not a good idea either, in my book.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum