Re: If not Ada, what else...

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Sat, 23 Apr 2011 16:30:42 +0200, Georg Bauhaus wrote:

> On 4/23/11 2:23 PM, Dmitry A. Kazakov wrote:
>> On Sat, 23 Apr 2011 04:50:43 -0700 (PDT), Piotr Trojanek wrote:
>>
>>> The question is what is your focus in a robotics research? If it is a RT control,
>>> than I see no point in not using Ada. If you are interested in AI, then Lisp is
>>> probably the way to go.
>>
>> I always did AI in Ada and see no point in using Lisp for that. There is a
>> common opinion that Lisp and Prolog are good for AI, which is as ungrounded
>> as the opinion that C is good for drivers. I think that Ada is especially
>> good for AI because AI is quite nasty when it comes to testing. Often you
>> cannot figure out whether the problem is due to a bug or a logical error.
>> Ada sufficiently reduces the bug rate. That helps much.
> 
> Would the work have been just a little easier if there
> was a plugable standard garbage collector?

No, because I doubt that a standard collector would work with the
heuristics I used, e.g. for factorization of decision trees, pruning,
cloning upon modification etc. GC does not help in establishing various
"uses", "depends" relationships, like strong, weak, read-only,
weakly-ordered references etc. Which is the whole work. These relationships
cannot be derived from the program, they must be programmed. Since you have
to do this anyway, collection becomes a negligible problem. I don't
remember if I even had a memory leak.

I am very suspicious about usefulness of standard GC in general. I think
that Ada represents a good example of why GC is not needed. If Ada
supported unconstrained components in a less boring form, I bet that the
issue of GC support could be closed. I also think that semantically the
idea of magically working GC contradicts to higher level abstractions. It
can work only in some strata, not too high, not too low. Hence GC is bad
for design.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de