>> Nathanael was planning to throw away SystemOrganization and write a 
>> new one for traits so there is an opportunity there. The problems is 
>> that we cannot invent the problem with file based system if we do not 
>> have to do it. So we will see what will happen and if the new 
>> SystemOrganization is capable or not of that and how other can adapt 
>> it to their needs.
>
> I would guess the important things would be:
>
> - make SystemOrganization a well defined interface, and build in the 
> assumption that there may be multiple implementations
> - don't assume that any of these implementations are hanging onto real 
> class objects or CompiledMethods
>
> As long as you kept to that, I doubt we would run into too many 
> problems.
>
> Avi

Then this is perfect. We will try to come up with something good.

Stef