> The main problem we have encountered with Squeak is the dispersion of
> knowledge and change files needed to get anything to work.  From a
> commercial perspective, we need to reliably be able to 1) build the VM 
> for
> each platform from C to maintain control and reliability, and 2) have a 
> well
> defined change set load that works on each VM.  That said, we are 
> getting
> close on #2 and haven't started #1. It seems like the knowledge base is
> fragile, not the main code base.
>
Building the VM really ought not present any problems, especially now 
VMMaker is in the main image.

With some minor wrinkles still to sort out for each platform (getting 
used to where the makefiles are and what to do with them is still my 
bugaboo) it really comes down to:-
Grab all the handwritten C files, makefiles, resource files etc from 
SourceForge. Take a look on the swiki for the pages decribing how to do 
this - and offer edits for any parts you find confusing or inadequate!
Make sure you have an up to date image and the VMMaker is in the 
system - if you _have_ to use an older image, I can probably dig out 
VMMaker changesets suitable for realtively old ones.
Run VMMaker. There are two ways to do this, one via the reasonably nice 
UI (VMMakerTool openInWorld) and the other via simple scripts (which 
might be more useful  if you want to automate the process. you can even 
generate the source for all your platforms from the one machine by 
telling VMMaker to generate for the other platforms)
Run the makefiles per platform requirements.

By and large you should find that the VM does not dramatically affect 
which  changesets you can or should use. We do quite well at maintaining 
both forward and backward compatabilty in general.

And of course, if you're doing commercial work and need help with any of 
these things, do please wave some money at me; I could do with getting 
paid  sometime soon.

tim at sumeru.stanford.edu