Re: Is Ada likely to survive ?

[Brian Rogoff <bpr@shellx.best.com>]

On 2 Aug 1997, Robert Dewar wrote:
> Brian says
> 
> <<The fact that few seem to be working on extending Ada is certainly
> unfortunate IMO. Perhaps as we develop more experience with GNAT the
> situation will change. But I think the comparison here isn't completely
> fair. How much creative energy is going into extending the ISO Prolog
> standard, for example, compared to that going into extending Ada? ;-)
> >>
> 
> I strongly disagree. 

I disagree that you disagree, since I agree with the rest of your post.

> Putting lots of work into extending Ada (at the 
> language level) is not the most effective use of time. It is of course
> entertaining to do, and one would hope that, particularly in academic
> circles, there are those willing to play with language extensions
> (hopefully using GNAT to prototyp them).

If you are a programming language researcher, extending an existing
language with new features may very well be an effective use of time. 
I see a lot of work done using C++ as a testbed for new features, and I'd 
prefer that Ada were used instead. Of course, most of us are not language 
researchers or designers, so...

> But in terms of serious use of the language, frequent extension and adding
> of features is not very helpful. Far better is to work on the infrastructure,
> i.e. bindings, reusable libraries, the knolwedge base of how to use Ada
> effectively etc. Occasionally new pragmas and attributes may be useful,
> but full fledged language changes and extensions carry a heavy burden of
> proof to be worth considering.

I strongly agree with this. However, the accumulation of practical
knowledge from the language user community will suggest areas for 
changes, and the best way to test changes and guarantee that
they are really improvements is to write code in the extended language. A
good example is the "withing problem" thread; there seems to be consensus
that the workarounds are not always practical, but no consensus as to the
right solution. Some experience doing some limited "playing" with extended
Ada would provide valuable feedback for the next generation of Ada. I'd 
personally like to have *some* amount of type inference or automatic
instantiation of generics, though I am probably in the minority on this.

> Sure there are lots of people playing with ways to extend Prolog (it is
> very weak in some areas, and sure needs extension -- I still remember 
> the presenter at the fifth generation conference in Tokyo saying that
> they were looking at prolog, and it looked very good except in the areas
> of modularity and abstraction, but never mind, they would add these features.
> TO my taste, these "features" still have not been effectively "added" to
> Prolog (actually the idea of modularity and abstraction as add on features
> strikes me, and struck me at the time in tokyo, as a big ludicrous.)

Actually, some people in the ML community are working on just that idea, 
a "modular module system", which is added on to arbitrary typed base
languages. Of course, this may just confirm your suspicion that the idea
is ludicrous!

-- Brian