Some GNAT history (was Re: SGI GNAT Question? (Long))

[dewar@gnat.com]

Incidentally a bit of history may provide useful
perspective here. I know that many readers are fully
familiar with this, but we always have new readers who
have discovered Ada more recently who don't always know
the history (and of course it is great to see new people
coming into the Ada fold).

I personally have fought the fight to get open source
software out for Ada from the start, well starting in the
early 80's.

At NYU, we wrote NYU Ada/Ed, the first validated
Ada translator first in SETL, and then
later translated into C.

We had a big struggle to get
the sources of this out. The DoD which was funding
as at the time retained some data rights and wanted
to try to make money (more accurately to get some of
the money they had spent back) by selling the program
through the national technical information service.

Eventually we did succeed in releasing the sources, but
it was really too late to be useful.

Later in the 80's I made several trips down to Washington
trying to interest DARPA in funding a free software Ada 83
compiler. I got some expression of interest, but basically
DARPA was not that interested in Ada, and things never got
anywhere.

I considered it crucial for Ada's long term success that
we have open source versions available for use in
universities, but until Ada 95 came along I just could not
convince anyone else that this was important.

I spread the gospel of the importance of a free software
approach to Ada 95 to Chris Anderson, and from my input,
and from the input of many others, it was Chris who was
finally persuaded that this made sense.

The contract between the DoD and NYU for GNAT was rather
amazing. Following my specific recommendations, it required
the use of the GPL and LGPL (and indeed the entire text of
the GPL is in the contract), and it also requires
assignment of the copyright to FSF (I explained at the time
why I thought it was important to get the copyright free of
NYU, since some people at NYU, as at any other university,
are more interested in making money off patents and
copyrighted software than in promulgating free software.
I knew that an assignment to the FSF would protect the
continued free status, whatever NYU decided it wanted to
do.)

That's a rather *amazing* contract, and in retrospect,
Chris was surprised that she succeeded in getting it past
the legal and procurement scrutiny in the DoD. In fact I
think it is a tremendous achievment, and I am not sure
anyone else other than Chris could have managed it.

The next big hurdle was the other Ada vendors. Chris
thought they would be pleased at this development since it
was clear to her it would aid the long term goals of Ada,
but in fact several vendors went ballistic, since they
felt that GNAT threatened their markets. Note that not all
vendors felt this way, but some big important ones did,
enough to make a serious problem (I will not name names
at this stage, it is water under the bridge after all!)

The whole project hung in the balance, but squeaked through
by taking two steps:

  1) eliminating validation

  2) adding some siderals to the contract that eliminated
     funding for several features, enough so that the
     result could be seen to be crippled (the list included
     fixed point and subunits, and some other stuff I have
     forgotten).

So things forged ahead. The validation did not really
matter, although it is ironic that when it came to awarding
the contract for the academic compiler, the main reason
that Mike Feldman's proposal for a GNAT based solution was
turned down was the lack of validation! He had a letter of
commitment from ACT which existed by that time, of the
intent to validate, but was told that ACT was not a
sufficiently credible company to ensure validation -- a bit
ironic as ACT is still the only company to have achieved
full 100% validations of the core and all annexes :-)

(by the way Mike [Feldman] if you read this and want to
correct or elaborate, feel free!)

As for the siderals, we found a way around this. Both I and
Ed Schonberg had sabbaticals due at NYU, and we took them
during the project. Instead of rushing off to some far away
land and thinking high academic thoughts, as is often done
during sabbaticals, we stayed at NYU, and worked on the
missing features. Since NYU, and not the DoD, was paying
our salaries, no one could complain that DoD money was
being used for prohibited features, and that is how GNAT
came to be 100% complete, despite the insistence of other
vendors in these siderals in the contract!

The GNAT contract ended after four years (the total by the
way was about $3 million during this period, a small
fraction of the total money that the DoD had spent in
direct and indirect subsidy of other Ada efforts
previously).

That's when ACT was formed. The cast of characters was
pretty much unchanged from the NYU days, and in many
respects the spirit of the project continued unchanged.

The three changes were

  1) The acquisition of serious paying customers, with SGI
  playing a crucial role in embracing GNAT and its open
  source basis early on.

  2) A serious commitment to validation, which had been
  prohibited during the period of the NYU contract.

  3) A considerable increase in the scale of the project,
  supported by these paying customers. At NYU, we had
  the equivalent of 6-8 full time people, and ACT and
  ACTE now have more than double this number.

As a result, development has continued without any more
expenditure of tax payers money, and a nice synergy has
arisen between the paying customers, who get a supported
commercial product that they can use in major projects,
and the users of the public version, who provide a large
testing community, and help smoke out problems and also to
suggest improvements, many of which suggestions have been
adopted in successive versions of GNAT.

We have continued to maintain the practice started at NYU
of building public binary versions. Although of course the
sources are fully available, for many users it is important
to have prebuilt binary versions, hopefully nicely packaged
for easy installation. Building and preparing these public
release versions is certainly not zero effort, but they
have had a significant role in increasing the spread and
use of Ada, a goal that ACT (and I personally) is (am)
committed to pursuing.

Robert Dewar






-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own