Hi,

The last days I was thinking in some problems we have.  The problems that 
are walking in in my head are basically 2:

- The impossibility to find a good way to evolve the monolithic image to a 
modular one. - The problems I found trying to get absolute transparency in 
rST. What I means is: The problems to evolve a monolithic image to a 
distributed one.

Thinking in a solution to these problems I remembered an Alan's phrase: 
"Extreme Late Binding".

Since Smalltalk/80 we didn't find the way to add more late binding to 
Smalltalk.

For example, today we have (most of the times) hard-coupling to Classes 
(using the globals).  Also we have hard-coupling to some "especial" objects 
as nil, true, false, SmallIntegers, etc.

The other source of hard-coupling is the extensive use of identity.  When 
you use identity is not so easy to change the object (examples: #isNil, 
#ifTrue:, #==, etc).

Yes I known what you think: realize of a problem is the easy part, to find 
a solution is a *really* different history.


I have not the answer, only some vague ideas...

What we need, IMHO, is some mechanism to refer to objects based not on 
names (variables, literals) but based on some characteristics of our target 
Object.

I'd like to hear your comments,

Diego Gomez Deck