Sat Nov 14 10:20:54 CET 2009
The current implementation isn't so efficient for memory usage.
From I gather that a one-size mark/sweep algorithm might be
better. It would also remove the need for restarts. However, it
would complicate the use of variable length objects.
Additionally, pairs are represented quite inefficiently: 3 word cells
per pair. However, in typical applications, the amount of memory used
for representing cells is probably much less that used for
representing leaf objects. Let' s not worry too much about this..