Re: [RFC] Header organization of -base & -gui

[David Ayers]

Adam Fedor wrote:

> On Thursday, July 10, 2003, at 05:49 AM, David Ayers wrote:
>
> > Well this is a decision that just has to be made explicit before we 
> > can continue...
> >
> > a) We leave the gnustep/ directory (and also leave gnustep/base, 
> > gnustep/gui) and install Foudation (when applicable) at top level 
> > (well, with --disable-flattened under the library combo) and require 
> > -make to install gnustep.
> > b) We move everything to the the top level and support custom OS X / 
> > PB projects.
> >
> > I personally have no preference.  Yet, I would prefer a final 
> > decision soon though.  And we should at least inform -discuss about 
> > the decision before we commit anything affecting this.
>
> I'd rather not add too many top-level header directories. But one of 
> the points of these changes was to support packages like frameworks. 
> So we should probably do (b) but be picky about what we want to 
> install... 

So, you are saying that we should keep the Headers/gnustep directory but 
just not install gnustep/gui and gnustep/base?  But we still should move 
unicode as it's part of -baseadd, right?  Well, as long as we're sure 
that other projects won't be moving I guess I could live with it.  I 
just wouldn't like to move headers again in following release cycles.  
(Maybe we do need more time to fiure this out.)

> Yes. The back headers were installed when the API was young and some 
> developers needed them to get access to X Windows. I don't think 
> that's necessary anymore except in the rare case where some one is 
> trying to implement something very backend specific and the API to 
> handle it isn't defined. 

Can I interpret this as an "OK not to install"?  I think the /rare 
cases/ can rely on private local declarations.

> Doesn't matter to me. The Foundation Extensions is completely 
> unsupported (as far as I am concerned) and technically shouldn't even 
> be in the repository (wrong license). I've kept it there because the 
> db and gsweb library needed it. Hopefully we can transition away from it.


I'll work up a patch for db to use -baseadd instead, while waiting on 
the "the decision".  Richard already fixed gsweb.

Cheers,
David