Re: Ada and pointers

["Marin David Condic" <dont.bother.mcondic.auntie.spam@[acm.org>]

I suppose that my logical explanation doesn't necessarily cover the same
territory as mathematical proof. Obviously not all algorithms have bounded
memory use - try to compute Pi to the last digit - you'll run out of memory
before you get there. I'm not sure if thats what you mean.

From just a logical perspective: All computers in the known universe have a
finite quantity of memory. Hence all dynamic allocation, on all known, real
world computers is a fiction built on top of a static allocation that is
determined at computer-build time. So I'd suggest that it stands to reason
that if you have a dynamic algorithm that actually runs and computes a
solution on some real world computer, then there must be a static allocation
solution that will run on that same computer.

Now maybe we should spend some time figuring out how to trisect an angle? It
beats chasing down C program bugs which is what I have on the agenda for
this afternoon... :-)

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Hambut" <hfrumblefoot@yahoo.com> wrote in message
news:fb75c450.0108151054.74fdfc5@posting.google.com...
>
> Well the statement that pointers can be mimic-ed through other
> mechanisms is certainly true enough, and I agree with your other
> points but........
>
> I was thinking about the dynamic allocation aspects - if you can't or
> don't want to bound the memory used then it becomes difficult to see
> how you could work without dynamic memory allocation (which is,
> admittedly, not precisely related to pointers - but they provide a
> popular way of getting access to dynamically allocated memory).
>
> Of course this is all fairly non-specific, but in terms of a formal
> mathematical proof - well (and here's a leap into the unknown coupled
> with thinking aloud, so feel free to flame the rest of this...)
> wouldn't this mean that you'd somehow have to prove that all
> algorithms had bounded memory use?  This seems a hard thing to do (and
> I'm not precisely sure this would completely prove that pointers can
> be disregarded - for example what about alias-ing capabilities?)
>
> That said I agree that in a practical sense you can probably do
> without pointers, albeit with some effort and discomfort.
>