Re: gnat-3.10

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

Cory says

<<The GPL never states that you have to make a public release or you
have to give anything to anybody.  It states that if you give someone
a binary release, you have to give them source (which ACT does, the
source is available to customers).  It also says that you cannot
restrict what someone does with what you give them (as long as they
comply with GPL), so ACT asks customers not to release the sources,
but it really cannot require them to.
>>

To clarify here, we tell customers that we think it is in the best
interests of GNAT if prereleases and wavefronts are not released
generally (for all the reasons I have previously stated), but that is
absolutely right, we cannot require it.

This is a similar situation to the GCC snapshots, which are available on
a limited basis, with a similar note that it is in the best interests
of the GNU project if these snapshots are not distributed. But there too,
it cannot be required, and in fact at least on one occasion, a public
release of Linux on CD/ROM has included the snapshot sources, much to
the consternation of FSF and Richard Stallman.

Richard Stallman is in fact quite insistent on the undesirability of
releasing snapshots generally. His concern is just like ours. It is
one thing to have a single user run into difficulties under conditions
where these difficulties can be immediately addressed. It is quite another

to have a public release where there are problems. In the latter case,
there will be general yelling and complaining, and this does not help

the GCC community at all.

I will reiterate that our release policies are geared to the needs of
the general GNAT consuming public, which is a very large set of people
at this stage, most of whom have very little expertise in getting arond
problems (you would be amazed at the number of reports of installation
problems we have that arise simply from not carefully following the
installation instructions).

Yes, there are a few enthusiasts who would like to fiddle with the
absolutely latest version, but we do not gear our public release
policies to this small group!

P.S. When I mention Richard Stallman's insistence on the undesirability
of releasing snapshots, I am reminded of an incident a while ago in which
we discussed with him the possibility of releasing snapshots with one
public version of GNAT. He got very agitated, and yelled at us (this
was in person, not by email), and we quite understood this reaction :-)

Note that the issue of figuring out appropriate release dates is one that
all vendors of software face. The fact that software is free software
really makes no difference at all. There are people who would love to get
their hands on Windows 97 right now, but Microsoft has decided it is not
ready yet.

Sure, vendors may make mistakes, but the temptation is in the direction
of releasing too early. Just the other day, I suggested an early public
release of 3.10 to the team here, but the discussion overruled the
suggestion, since people pointed out several points (including the issue
of dragging in the tasking stuff all the time) that really need solving
before the public release.

You should understand that I am eager to get 3.10 out as early as possible,
for one thing, I would like people to be able to use the super-neat SPITBOL
stuff that I have added. But it is a long-term mistake to let this
eagerness result in a premature release.

Robert Dewar
Ada Core Technologies