Re: Government Policy on Ada Acquisitions

[dennison@romulus23.DAB.GE.COM (Ted Dennison)]

In article <1994Sep7.155252.14027@ocsystems.com>, kdh@ocsystems.com (Kevin D. Heatwole) writes:
|> Jim Dorman (aetechca@powergrid.electriciti.com ) writes:
|> > 
|> > The way to ruin a budding commercial enterprise is for the 
|> > government to PAY NON-PROFIT ORGANIZATIONS TO DEVELOP THINGS AND 
|> > THEN GIVE THEM AWAY. For the benefit of those in academia, this is 
|> > how things are actually working. The problem with this approach is 
|> > that it ruins the incentive of industry  to compete. Why should 
|> > Alsys, AETECH or RRS commit $900K to develop and maintain a 
|> > validated $99 Ada9X compiler for academia, when the government 
|> > issues a non-competitive sole-source contract to NYU for $2.5 
|> > million dollars to build a non-validated GNAT compiler from 
|> > scratch to be given away free? 
|> > 
|> 
|> Being an employee of one of the "budding commercial enterprises" in 
|> the Ada industry, I, too, share Jim's concerns (I'd like to remain
|> employed ;-)).  The costs of competing in this environment are very 
|> high (you need a minimum of at least $1 million a year in on-going 
|> revenue if you are even going to think about supporting an Ada 
|> compiler).  That being said, I am still confident in our ability to 
|> compete with any "cheap" compiler, even if it is "of high quality", 
|> but it does make you think twice before entering the game.

I wouldn't worry too much. I implemented a program in Sun (Verdix) Ada 
and then translated it to GNAT (our customer site has no liscence for
Sun Ada). GNAT's executable ran about %20 slower. Any company that 
needs a fast product should be willing to pay a reasonable amount for
a commercial compiler. 

Of course, if your commercial compiler costs more than a >20% faster 
CPU you may still have a problem.

For you legal types, the program made heavy use of floating-point 
math and very heavy use of disk I/O. Neither executable was compiled
with optimizations.

|> One solution might have been to build into the GNAT contract with
|> the government, the restriction that GNAT could not be validated for
|> a specified period (say, until the year 2000), even by an independent 

If this is your worry, worry not. My understanding from reading the gnat
docs is that they NEVER intend to validate gnat.