> I don't have specifics to add to this, but I wanted to write that I
> think what you and Andrew Brault and others are doing with
> PocketSmalltalk is fantastic.
>   http://www.pocketsmalltalk.com/
> And now, integrating the concepts with Squeak. Incredible!

Thank you :-)

> This sounds fantastic if I understand it -- basically you are using one
> set of Smalltalk classes (Squeak's) to browse another set of Smalltalk
> classes (Pocket Smalltalk + App). This is also a nice step in the
> direction of firewalls (where one VM talks to another running in a
> different objectspace).

Its not really in that direction.  In fact, until the code is packaged as an
application, it is compiled by the Squeak compiler.  When code is executed,
it is executed by the Squeak VM.  So in effect, one is starting from a VERY
minimal set of code, and it can be grown to build another Squeak, if you so
wish.  I don't think that that is necessarily a good idea, but each his own.

This will allow developers to run the Pocket Smalltalk code from within the
Squeak environment for testing purposes.  Now remember, that Pocket
Smalltalk assumes the ability to call OS calls directly.  These will have to
be emulated to make this work well.  When they are ready to package, then
the packager will attach the Pocket ST vm and code together into an
application.

> Maybe in time you will be able to generate images using the
> PocketSmalltalk classes that run on the Squeak VM? I think there would
> be much value to this, and it might then be possible to build up a
> complex Squeak image by addition of Squeak code to the PocketSmalltalk
> base rather than strip an image as is currently done. This then might
> help eventually to address the very same modularity issue you raise
> here.

See above.  I think that this is very ambitious in a lot of ways.  Althought
I would agree that Squeak has everything and the kitchen sink thrown in.  It
has MANY features that have grown out of lots of work that would be
difficult to redo from scratch.  I think that with a decent amount of work
you could get modules and Squeak to shrink to a decent size.  My problem is
that I need more than just a decent size, I need something that will run on
ridiculously small size, which is why Pocket ST has diverged a little from
Squeak's approach.  I believe that different design decisions must be made
for the smaller space, than would be for the types of deployable
applications would like to do.