[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
Garbage Collection Summary
- To: <xanatech>
- Subject: Garbage Collection Summary
- From: Marc Stiegler <marcs>
- Date: Tue, 21 Nov 89 13:44:25 PST
Reading MarkM's recent discussion of garbage collection, C++
compiling, etc., I found it, uh, somewhat difficult to see the
forest behind the trees. So here's my stab at a brief, high-level
summary, for others on this channel who suspect that something
very important was under discussion, but aren't quite sure what
it was:
1) Our garbage collector imposes some peculiar constraints on
programming style until C++ compilers are fixed to match the
C++ specification. Depending on individual temperament, different
people will view the peculiar constraints as infuriating and
intolerable, not merely peculiar. Surely, if we were to force
third party developers to live within these constraints, we would
receive howls of protest from a group of significant size.
2) Fortunately, we are not going to force these constraints on
our third parties.
3)Developers who are using C will never see any of this at all--they
will interface in a more direct manner to the backend. An independent
though related question is, to what extent will C developers
be able to exploit our tools (above and beyond the backend itself)?
Bob Perez is addressing this question (though at a limited priority
until he has the tools documented at least for C++ users :-)
4) Developers using C++, who want to use tools above and beyond
the backend itself, will probably never see it either. We are
investigating 2 paths to eliminate the constraints:
a) Leverage outside resources to fix the C++ compilers, and
b) use the technology inside our existing C++ preprocessors
to remove the constraint before the compiler sees the code.
5) Both approaches to fixing the problem have only a small impact
on our critical path (on the PERT chart, MarkM's calendar is
custom-built to automatically account for occasional events like
this).
6) MarkM's second message suggests a third alternative, building
a less effective but more convenient garbage collector for the
third parties. This would fall under the task of building extended
development environments for third parties, to be addressed when
we get seriously into the toolkit development discussed in Joel's
marketing plan.
--marcs