[gclist] Fragmentation issues

[Stefan Monnier]

> My take on it is that fragmentation is mostly a myth---given a good allocator,
> most programs don't give much fragmentation.  So non-copying collection is
> much better than people generally think.

I don't really care whether fragmentation is high or low, but I'm wondering
about its evolution in time. How does it deal with the case where the program
is to run for months without interruption ?

How does Cedar deal with this ? (does it compact memory from time to time using
something akin to the mostly-copying GC or does fragmentation problems never
show up early enough to be noticed ?)


	Stefan