Re: Reasons, why Miss Scheme should collect her garbage before shutdown.

[Matthew Flatt <mflatt@cs.utah.edu>]

Quoting Sebastian H Seidel:
> Thus, I'd like to try out the
> more precise GC that goes by the name of "3m". Does it include a
> copying facility to avoid fragmentation of the heap? 

Yes.

> Maybe it's not
> yet ported to Windows anyway? 

MzScheme3m should work in Windows. (Actually, I haven't tried it in a
couiple of months. I'll have to double-check that nothing's broken.)
MrEd3m doesn't, yet.

> If I'd program an extension using "3m":
> would the interface to the GC remain the same? Does it interact well
> with embedding and extending C-programs?

This is the main problem with 3m in its current form; the interface for
extensions is not even documented, because it's too difficult to use.

I plan to re-arrange the interface sometime in the next few months, and
the result will look much like the Scheme48 interface, where all C
pointer variables (glocal and local) must be registered explicitly with
the collector. But the allocation and finalization interface stays the
same.

In any case, 3m will never be as convenient as the conservative version
for embedding or extension, so I imagine a future with both.

> Maybe 3m" will even be the
> default GC of PLT-Scheme in the future?

Yes. It's difficult to say how far in the future, but I'd
optimisitically guess one or two years from now.

Matthew