Re: Next Ada compiler for Debian: the votes so far

["Steve Whalen" <SteveWhalen001@hotmail.com>]

Thanks for this summary.  I think these are by far the most important
issues to be included in the Debian log (or whatever it's called).

It would be nice if a sponsor could be found to pay AdaCore to do GMGPL
releases periodically, but I'm starting to wonder if AdaCore/ACT always
intended to "go proprietary" as soon as they could. With as "soon as
they could" being whenever another "version" of Ada was standardized
that was NOT called Ada95.

During the first 5 years or so of GNAT's life AdaCore seemed to
genuinely believe in the benefits of having a completely free compiler
available, at least they convinced me.  Of course they may just have
been saying it because the DOD was paying them to make a free compiler
available and so they had to.

It's my understanding that the GNAT Ada95 compiler was basically built
with DOD $$$ on the condition that it be free for all uses and users
(but ACT and others were free to charge for support as long as a
compiler remained free). This specifically included the libraries being
under a LGPL / GMGPL type license.

If the Ada2005 compiler from ACT is built primarily on the Ada95
compiler (as opposed to a complete rewrite of the compiler code and
support tools) then I think AdaCore is coming pretty close to violating
the spirt (but I'm sure not a lawyer's reading) of the DOD contract
that created GNAT.

By "closing" off the creation of commercial applications with the GPL'd
Ada2005 compiler the Ada2005 compiler would no longer meet that
requirement of the original Ada95 DOD compiler contract.

I hate to see this happen to what I think is one of the best decisions
made by our Department of Defense (creating the free Ada95 compiler to
help offset the removal of the "Ada mandate").  What AdaCore is doing
certainly seems to violate the spirit of the contract that enabled the
creation of GNAT (unless of course the Ada2005 compiler uses none of
the work product created under the DOD Ada95 contract in the Ada2005
compiler: my understanding is that Ada2005 contains a lot of Ada95 code
but I haven't compared them).

The bottom line is that neither GNAT nor AdaCore would exist without
that DOD contract for Ada95.  Something sure has changed: here's a
lengthy but relevant quote from Robert Dewar from about 10 years ago
(for anyone who doesn't know who he is, I'll keep it simple and call
him one of the principle architects and driving forces behind the GNAT
compilers up through at least release 3.xx and probably beyond):

(quoting from comp.lang.ada:

Robert Dewar	  Aug 2 1995, 12:00 am)

...

For example, the contract between NYU and the Federal Government
requires that the compiler be distributed uner the GPL and LGPL.
Suppose instead that it had been distributed under the public domain.

In that case, Ada Core Technologies (or any other company, but this
most often happens with the original authors) could take the existing
PD product, continue to develop it, market it at whatever price the
market would bear, and treat it as a fully proprietary product.

This would presumably be fine for those willing to pay big $$$ for Ada
products, and might indeed make the GNAT authors richer. This scenario
has certainly played itself out with other products that were initially
in the public domain.

The trouble is that then, although the public domain version continues
to exist, it languishes, and pretty soon, is left in the dust by the
commercial version, and people find themselves forced to pay high
prices or left in the cold.

The nice thing about the use of the GPL is that it entirely prevents
this from happening. If someone pays SGI, or ACT, or any other company
to enhance and maintain GNAT, then all the improvements and changes are
guaranteed to be available to those who need an Ada compiler to work
with but do not need commercial level support. This is a much better
model for the whole community I think.

Yes, of course it is the case that other Ada vendors would prefer to be
able to grab public domain stuff without being encumbered by the
restrictions of the GPL, so of course they would prefer that all the
volunteers out there provide them with free code they can use. That's
perfectly understandable.

Now if this continued fredom were achieved at the expense of seriously
limiting use of the resulting software, that would be an unfortunate
side effect. In a world with no proprietary software, they there would
be no issue. In recognition of a more practical world in which free
software and proprietary software continue to co-exist, the modules of
GNU C and G++ are distributed with four different levels of licensing
ranging from the GPL, to completely unrestricted, with two other levels
in between. The status of each module is chosen to achieve an
appropriate balance between trying to maintain the free availability of
the software and trying to maintain its reasonably unrestricted use.
This has been quite comfortably achieved with GNU C, which has been
used by a wide variety of users in a wide variety of commercial and
technical settings, and there is no reason to think that GNAT will not
be able to achieve the same appropriate balance.

(end of quote)

I think AdaCore/ACT has lost that "appropriate balance".

Steve