Re: what DOES the GPL really say?

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Les says

<<No, but usually the discussion is in the context of whether you have
to distribute changes at all, which of course is not required.  However
this time the issue seems to be about distributing among a small
circle of friends with at least an implicit agreement that none will
distribute additional copies.  I fail to see any philosophical difference
between this and selling copies with a contractual requirement not
to redistribute.
>>

Effective software development requires that the scope of use and testing
of a piece of software be commensurate with the stage of maturity of the
project.

When a fix is first developed, it remains in the domain of the fixer until
he or she is reasonably confident that it is a correct fix.

Then it is propagated to the development team and other people try it
out in the context of other things that are going on.

When the team collectively thinks that a set of fixes is ready, then it
needs to go outside the team to a small selected group of users to see
if things work (a process usually called alpha or beta testing depending
on the perceived maturity of the release, really I am talking about what
I would call external alpha testing here).

Then when a release seems to be in a potentially releasable state, it
needs to be more widely tested (this is a typical beta test situation).

A product can then be initially released to a wider circle of customers,
and eventually a full release can be made.

If the requirements of the GPL made it impossible to follow this procedure,
then the GPL would in practice make it very difficult to develop reliable
software.

But the GPL in practice does NOT make it impossible to follow this
procedure. Although further distribution is always permitted, during
the early stages, those involved understand that it is not helpful to
widely distribute that which is not ready to be widely distributed.

In the case of GCC, the snapshots are available to a number of people,
but the Free Software Foundation notes that wide distribution of these
snapshopts is not helpful to the GNU project, which seems an exactly
accurate assessment, and since people perceive that it is an appropriate
assessment, they generally agree. There are some glitches, at one point
a CD ROM of the Linux distribution included the snapshots, which was
potentially highly disruptive, and some loud yellling at the people
involved went on to try to avoid this happening in the future.

Note that people's willingness to abide by the informal "do not
distribute" rules is based on their perception that the request is
reasonable, and is indeed based on a concern to avoid premature
distribution. If they felt that the request was based on a desire
to hoard software that was in fact ready for wide distribution, they
would not concur, and the redistribution would occur.

There is a huge philosophical difference between seeing distribution
restricted because a group of people agree that it is not a good idea
for the good of the project involved to distribute software prematurely,
and a situation in which distributable software is being hoarded.

Richard Stallman is adamant about the importance of free and wide
distribution of software as I think everyone knows, but he is equally
adamant that it is a bad idea for insuffiently tested snapshots to be
prematurely distributed widely, and will yell loudly at anyone who
"misbehaves" in this manner. I know that, I have been yelled at!

In the case of GNAT, our position is quite simple, our beta testing
works by having very small sets of customers get wavefront releases.
They are willing to test these out, because these beta versions fix
problems they need to get around, and they are willing to risk the
instability of a new version to get around their problem. In my
previous message, I outlined the internal QA testing we do to try
to minimize the possibilities of such instabilities.

When we have something that we feel is potentially ready for release,
we distribute it more widely to our supported customers. And more of
them pick it up. That actually is a relatively new procedure for us,
and resulted recently in the internal prerelease designated 4.10a.

Well what's the result? In fact 4.10a is looking good, BUT there are
a couple of glitches that have not proved a problem for our customers
except in one or two cases, but which absolutely MUST be fixed before
a full public release of the 3.10 (oops I mean 3.10 when I mention
4.10 anywhere above). We are now working towards a 3.10b. We will
make this available to our customers again, and if it is glitch
free, then release it publicly.

As I noted in a previous post, releasing free software is actually
more difficult than releasing proprietary software. In the latter case,
the user community is self limiting to those who are willingf to shell
out the money, you have a chance to track the users, and most importantly,
you provide at least some kind of support.

In the case of free software releases of GNAT made by ACT, the new software
is immediately used by tens of thousands of users, we have no idea who is
using it, and we cannot provide support to those who are using it.

That means that we have to be SUPER sure, to the best of our abilities,
that the release is glitch free. We are determined to work carefully
towards this goal, and will not release 3.10 before we think it is 
ready, no matter how loud Ronald and others may yell. Looking back,
if we have made release mistakes it has been on the side of releasing
too early (in particular the annoying Constraint Error on syntax mistakes
in 3.04, and the loader problems in the 3.09 NT come to mind -- the
latter is actually instructive, we slightly rushed the 3.09 NT, to get
out a CD ROM for STC. Unfortunately, there is now way to cut the CD ROM
via the PAL without making the version public, so we were pushed into
releasing this a little earlier than we wanted. Now 3.09NT has proved
useful to many many people, but it has also caused some unnecessary
frustration).

It is our goal to avoid such problems in the future. If this annoys some
enthusiasts who are panting for the latest version, I am sorry, but I
think it is more important to get solid releases, than to get less solid
stuff earlier. 

Robert Dewar