"Withers, Robert" <rwithers at quallaby.com> is widely believed to have written:

> Awesome!  Thank you, Tansel.  This simply must make it into mainline squeak.
> If this gets delayed for 2 years, I'm gonna go bonkers.  I suppose it is up
> to SqF to make this happen.  Shall we shoot for May 1 for the 4.0 VM, just
> in time for SqueakEnd?
Well I'm certainly keen on getting it all into the general system -
after all the NewCompiledMethod stuff has been around for what, three
years now?

However, a big change needs thought. Changing the image format is a
thing we want to do very rarely - hopefully never again. There are a
number of things that need to be fitted into such a change that are not
directly part of Anthony's work, nor the NCM stuff.
For example, prim numbers are split into two fields in method headers;
it would be nice to fix that - but then with named primitives, maybe we
don't need any more than a hundred or so numbered prims.
For example, some plugins have awkward names (Klatt might be better as
KlattVoicePlugin?) or names that could simply be the same as the plugin
class.
For example, the table of conversions of numbered prim to named prim
could be dropped.
For example, we might consider HMMs  two-bit tags scheme again.

See what I mean?

In the meantime, anybody that has the time and interest should most
definitely bang on the code and do their utmost to break it in ways that
teach us how to improve it, make it as utterly wonderful as possible and
generally help Anthony.

Anybody tried out coroutining with it yet? A block return from one
Process to another? Fifty-deep nesting? Mutual recursion?

tim
-- 
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Stack Error:  Lost on a cluttered desk...