I'm interested in the plans that exist to improve the performance of
Squeak. 

For example, where between bytecodes and native machine code does
threaded code lie?

How are the translated methods managed?

Are there plans to incorporate run time type collection and inference
as in the Self project?

What makes Morphic slow (to the extent that this is known - I guess
if there was a precise and ready answer to that, it would be far less
so...)?

If the answers to these question are posted somewhere, point me in
the general direction please.
-----------
Daniel Vainsencher 
A member of the First Millennial Foundation,
making space a place to live in. Visit our CyberHabitat at
http://www.millennial.org