[gclist] What wrong with the GC in xemacs

[Tom Tromey]

Erik> I am not very well versed in GC.  do any of you know the Emacs
Erik> GC well enough to say something about how hard it would be to
Erik> replace it with, say, a generational collector?

Disclaimer: I'm not a GC whiz by any stretch.

But don't generational (and incremental, for that matter) GCs require
a write barrier?

If so, then modifying Emacs would be a big pain.  There is code all
over Emacs that looks like this:

	CDR(foo) = ...

Every such assignment would have to be found and modified -- the very
definition of "tedious" and "error-prone".

Anyway, as Jim Blandy says, any such work might be wasted.  If the
future of Emacs really is Guile, then you are probably better off
improving Guile.  Of course, if the time frame for this is measured in
years, then maybe improving Emacs's GC now would be worthwhile.

If I'm wrong about what it takes to implement these kinds of GCs,
please enlighten me.

Tom
-- 
tromey@cygnus.com                 Member, League for Programming Freedom