[self-interest] reflection (was: embedding based prototype languages)

[Jecel Assumpcao Jr]

On Saturday 09 November 2002 04:49, Brian T Rice wrote:
> [OOPSLA paper]
> This work looks very promising but requires some time to stop and
> freeze the objects (and the relevant protocols) involved.

They claim that for even very large type hierarchies this only takes a 
small fraction of a second on a fast machine. And that they have an 
incremental version of their system (not described in the paper, I 
think) where this would be even less of a problem.

> Possibly
> the best way to handle it would be to allow a collection of objects
> to use one meta-object to describe and implement that lookup
> algorithm for them all collectively.

This would work very well with their system if you have one meta-object 
per "type partition", but I think it is the overhead of grouping 
objects into partitions that you are trying to eliminate, right?

The numbers in the paper were very interesting, though I think that 
there are far more "types" and messages in Self 4 than what they 
evaluated. I wonder if that would have any effect in the results.

-- Jecel