Hi Stef,

> Yes but my point is can we find a clever way to use the
> same optimized way to support state extension. I do not
> like the idea to really on dictionary.

But you do like the idea of relying on an array, do you? ;-) Because that's
what you use today - an object is represented as an array that you access
via bytecodes which are the precise equivalent to #instVarAt: and
#instVarAt:put: ;-)

> Exactly. We were also thinking about that. Now the question is
> do you let or not this accessor be public or only accessible
> from within the class. The solution is... not simple.

Well, actually you don't have much of a choice since you can't really make
messages private. So saying it should only be accessible via the class is
unfeasable unless you do some rather drastic changes which violate the
simplicity of the proposal.

> We have been discussing that with nathanael and roel
> over the last year. But I think that not having iv would
> be a gain toward simplicity.

Maybe in terms of the algebraic notion of the system (as in Self) but I
really doubt that it would be simpler to explain and use such a system. As a
matter of fact that's why I chickened out of using implicit self (which
would solve this and other problems).

Cheers,
  - Andreas