Re: status of Ada STL?

[Brian Rogoff <bpr@bpr.best.vwh.net>]

On Tue, 18 Jun 2002, Robert A Duff wrote:
> Brian Rogoff <bpr@bpr.best.vwh.net> writes:
> >... but I think that Cyclone is
> > worthy of study, since it has some features contemplated for the next
> > generation of Ada (region based memory management, at least if I
> > understood Tucker Taft correctly) and also some features that would be
> > great to see in a new language derived from Ada (pattern matching).
>
> I was not impressed by Cyclone's region-based memory management.  At
> least for the kinds of programs I write, it seemed that most of the data
> would have to end up in the most-global (garbage collected) heap anyway.
> Regions are not first-class citizens in Cyclone (they can't be allocated
> dynamically, as in "new Region...").  But that restriction is key to
> Cyclone's ability to prove stuff about them (no dangling pointers).

That's right, Cyclone's implementors are from the ML community, so
being able to prove stuff, as you say, is a high priority.

> Tucker hates garbage collection, and longs for an efficient and safe
> alternative to both by-hand memory management and GC.

I don't know why anyone would hate garbage collection. That seems like a
rather odd sentiment. It also seems that if this is true he'll never be
satisfied, since GC is just the extreme on the automatic end for memory
management; all non-GC mem managers are "by hand" at some level.

In total half baked mode, my gut tells me that future type systems will
include a lot more usage information and compilers will be able to deduce
all sorts of nice things about your program so that even if there is no
"garbage collector" there will be little or no memory management explicit
in the code when you don't need to do it. But this is nothing like what
Ada or C++ or other low level bitdiddling languages are about, you'll need
to look at Clean's uniqueness typings or the effect systems research (by
the same guys who did the original region inference stuff I think).

> Me too, but Cyclone ain't it, IMHO.

Well, it ain't derived from Ada! Anyways, pattern matching, like all
modern functional languages have, would be a worthy addition to an
Ada like language.

> As you say, it's worthy of study; Tucker has mentioned some ideas for
> making it more powerful.

I must have missed this. Did the discussion happen here, or elsewhere?

> > http://www.research.att.com/projects/cyclone/
> >
> > yes, it would have been nicer if they'd started with Ada as a base, but
> > they're going for widespread applicability so C compatibility is an
> > important goal.
>
> C compatibility isn't the only thing.  There's also the fact that lots
> of people are used to C-like syntax.  Java isn't even close to
> compatible with C, but C programmers can feel comfortable with it,
> because it sort of looks like C -- I have no proof, but I suspect that's
> a big part of the reason Java caught on.

I disagree, and I think that there are lots of other reasons that Java
became popular, though I think that it's syntactic similarity to C helped,
I really doubt that was a large part of it. Why did Perl and Python (and
now Ruby?) achieve some popularity? My belief is that there is no great
theory of why languages become popular, and reasons that work in one time
and place won't work at others. It's just luck, or fashion.

-- Brian