Daniel R Risacher <risacher@worldnet.att.net> writes:

> While I understand and sympathize completely with ACT's position,
> (as stated by Robert) I suggest that the more commonly used model
> for GNU software is to have separate "stable" and "bleeding edge"
> releases.

No, this is rather uncommon, except for few examples.  In fact, *all*
of the GNU software distributed by the FSF tends to be distributed
mostly in what is considered a stable version (although typically some
earlier versions are kept for safety).  There is no explicit
distinction into stable and development versions for public releases.

>  Examples of this are Linux and the GIMP.  You cannot set up Linux
> 2.1.43 without seeing many notices warning that it is a
> developmental, kernel-hacker's version.  But you can download it
> just the same.  I wouldn't expect anyone to provide support for
> 2.1.43 without being paid for it.

Depends on what you call "support".  Actually, most of the support and
development in the various Linux usenet groups focuses around the
development versions, as there the most remains to be done.  If you
mean installation support by major Linux distributors, they will
obviously support what they provide, and that will indeed typically be
a "stable" version.

>  The GIMP is similar.  0.54 is billed as a "mostly-stable" version,
> but developers (and patient users) have been playing with 0.99.x for
> a while.  Support for 0.99 consists only of a mailing list of other
> users/developers.

And what superior support is offered for 0.54?

>  Guile has automatic, daily snapshots of the latest code available.
>  I suggest this sort of policy might be a reasonable model for gnat
> distribution.  I don't know; I don't use gnat, or Ada, for that
> matter.

It is usually quite a hassle to provide compilable, useful source
every day.  At the very least, you have to use something like CVS for
version control.  For some smaller projects this trouble might not at
all be worth the cost.

There are added disadvantages if you give development versions to the
public:  you have to sort out bug information and user foolishnesses
from sometimes inscrutible bug reports, and this can take more time
than you want to invest in it.  If you do alpha testing (giving out
preliminary versions to selected testers only), you can sometimes be
more productive.

So, in short, I think that there are different models that can be used
with justification.


-- 
David Kastrup                                     Phone: +49-234-700-5570
Email: dak@neuroinformatik.ruhr-uni-bochum.de       Fax: +49-234-709-4209
Institut f�r Neuroinformatik, Universit�tsstr. 150, 44780 Bochum, Germany