[ANSI-Smalltalk] code as standard

[Jecel Assumpcao Jr]

It has been suggested that actual code in the form of SUnit tests should
be part of the standard. This would be a very good start. It would be
interesting if each module could be divided into two parts: core and
extension. For the code we could have SUnit tests but for the extension
it would be nice to have actual code. This code would be defined
exclusively on top of the cores of its own and other modules as well as
extensions of indicated other modules. This would allow such code to be
loaded and run in any standard compatibly implementation.

How to achieve this without creating conflicts with equivalent classes
and methods already in the implementations is an interesting challenge.
And the code for the SUnit tests would have to depend on predefined
stuff or it would be impossible to bootstrap the whole thing.

Sharing code like this would move Smalltalk a bit back to the PARC days
and closer to the current single implementation languages (Perl, Python,
Ruby...) that seem to have more of an impact than standard based ones.

In case anyone is wondering if this is my secret plan to get people to
write free libraries that I can then use in my own implementation, I
very strongly prefer code I write myself to using other people's (I
don't even use operating systems or even processors designed by
others!). I have to fight this "not invented here" syndrome when it gets
in the way of something good, and in this case that something is a
possible unification of the Smalltalk community.

-- Jecel