----- Original Message ----- 
From: Peter Novak <pno at whitestein.com>
To: <>
Sent: Wednesday, December 01, 1999 12:59 PM
Subject: Re: SystemOrganization in 2.7


> Hello Stephen,
> my response will be maybe little bit off topic, still I think it can be
> relevant. In last few days we tried (Viktor Zigo and Co.) to implement
> our own namespaces. We used concept discussed two weeks ago. We put
> together Class and NameSpace ideas. So now Classes are NameSpaces for
> its inner classes. We were quite successfull with reorganizing Smalltalk
> system into same categories you've described (actually I was inspired by
> Your package browser ;>). Still we have some problems, 'couse we wanted
> to use late binding with the whole idea, so we have quite a big problem
> with recompilation of the system (exactly with expressions like A :=
> something - we want to use accessing methods rather than assignment, but
> existing code makes problems).

Instead of doing that, have you tried looking at what it would take so that assignments
to "variable bindings" are compiled to a #value: message sent to the appropriate association?
There really doesn't seem to be much effort involved in that, and it would only involve changing
code within the compiler framework.

les