Michael Rueger <m.rueger at acm.org> wrote:
> Daniel Vainsencher wrote:
> > Michael Rueger <m.rueger at acm.org> wrote:
> >>The ability of Ginsu and Monticello to deal with a semantic model 
> >>without loading the code first still needs to know where the code 
> >>originally came from. 
> > 
> > What? I didn't understand this at all. Please define "where the code
> > originally came from". Do you mean who the author is? or whether it is a
> 
> Which package and version etc. So if you load a generic st first and 
> then want to go to the packaged version you need to know which methods 
> belong to which package originally so can determine if anything has 
> changed in the meantime.
Ok. IIUC, you're talking about one of the possible solutions, where the
packaging information is put into an ST file so that the mcv can be
generated. Since I think most packages should be distributed directly as
mcvs, the few other cases don't warrant, IMO, changing the .st format.
It really is very easy to solve this by, for example, adding a class
comment in one of the classes saying "you can get the mcv for this code
at ...". For the few packages that really should have no load-prereqs,
this seems enough for now to me.

> > AFAIU, nobody is talking about changing the standard file out format.
> Actually we are. We need a way to attach attributes/semantic information 
> to the source code. Although it is possible with the current chunk 
> format it definitely isn't the best way to do it.
> All this of course is MHO.
> Well, maybe NSHO ;-)
What do you think IS the best way to do it?

Daniel