CLOS, MOP, AOP, SOP

Mats Nygren nygren at sics.se
Tue Jul 11 16:17:49 UTC 2000


Marcel Weiher <marcel at metaobject.com> wrote:
> [snip]
> 
> Interesting, an 'object-space' approach to backtracking / Prolog.   
> All the approaches I've are 'code-space', and that has probably  
> restricted my thinking in that regard, so I'd be glad to find out  
> more.  Anyway, are you aware of the bactracking-extension to  
> Smalltalk (without kernel support)?  It was an OOPSLA paper a while  
> back, and I put the code in Squeak last summer.

I have read it. I'm not sure its worth the trouble the way they did it.
To be really interesting it should be possible to backtrack on parts of
datastructures (composed objects:) and they didn't have a lot to say
about that. That is like the simpler "prologs" like turbo prolog that
didn't have "real" logical variables. I will come back to this after my
vacation.

Where is your code?

There is also the SOUL system to consider. Roel Woyts will help us
with that I am sure.

Another use for objects with history that I forgot to mention is as
a starting point for scripting. That is manipulate an object until
it becomes what it should be. Then look at the history and try
to generalize. In this way ordinary work and scripting becomes
to aspects of the same thing.

Yet another issue is the history alreay present in morphic. I would
like to relate to that but so far morphic has been a bit to much
unstructured code for my time constraints.

I will be happy continue this but beware that it will take some
time.

/Mats

PS. I have looked cursory on your trampolines, more about that later. DS.





More information about the Squeak-dev mailing list