Re: ARPA still undermining Ada

[carroll@stimpy.eecis.udel.edu (Mark C. Chu-Carroll)]

In article <37egd5$2vt@chnews.intel.com> kpriest@hopi.intel.com (Kevin Priest~) writes:
>
>I am unable to reconcile the following two positions:
>
>1. ARPA funds research, tries to take a long term view, and is
>interested in technology of the future.
>
>2. ARPA-funded research using Ada has been hindered because there were not 
>free tools available.
>
>Why did not ARPA simply fund the purchase of Ada tools, given their
>long-term view?  It seems incredibly short-sighted of a research
>organization to reject a superior technical solution because they would
>have to pay for the tools!  I'm sorry, Robert, but this defense of ARPA
>simply doesn't wash.

I don't see the difficulty.

When you start a research project, you generally look at what tools
you have available for that research project. That has a huge
influence.

To give an example, I'm a grad student working on a programming
language research project. When I started, we decided that we wanted
to work on parallel programming tools for general purpose parallel
programming. So we started looking for languages to use as a basis for
our work.

In our choice of a base language to work with, there were a number of
factors:

<1> We wanted a language that was actually used by people for
  general purpose programming - particularly using non-array based
  data structures.

<2> We wanted a language with strong typing.

<3> We wanted a languge with good modularity.

So far, it sounds like Ada would be an ideal choice, even if GNAT weren't
available.

But:

<4> We wanted a language where we had access to the source code for
  a compiler. We didn't want to have to rewrite all of the parsing
  and semantic analysis for the base language.

Now, even if you gave us a commercial compiler *for free*, we would
never have chosen Ada - because it would involve far too much effort
just to implement the analysis of the base language.

That kind of thing happens often in research - if you want to do
language work involving extensions of a base language, you really need
to have access to the source code of an implementation of the base
language. Before GNAT, there was no freely available source code to a
complete Ada compiler. And that, essentially, made Ada an unrealistic
choice for a lot of language work.

(As it turns out, we aren't using Ada. We discovered another research
project taking a *very* similar approach to ours, and decided that we
needed to change our direction slightly - so we decided to move more
into a more dynamic language, like CLOS.)

	<MC>
--