Re: Cocotron

[Richard Frith-Macdonald]

On 24 Dec 2006, at 02:00, Helge Hess wrote:

> On Dec 24, 2006, at 02:00, Yen-Ju Chen wrote:
> >  But I don't understand the difficulty of merging libFoundation and
> > GNUstep-base.
>
> Well, then just do it! :-) It doesn't make a lot of sense to argue  
> about it, if its trivial, just demonstrate it :-)

Fair comment ... though (unless you have already done so) I think you  
would need to assign copyright to the FSF before it would be worth  
addressing technical issues.

> In the past I have often posted the issues we encountered, but some  
> of the problematic areas are:
> a) KVC support
> b) FHS support
> c) release policies (aka no stable releases promoted)
> d) various minors
> e) packaging
>
> SOPE/OGo works perfectly fine on libFoundation *and* Cocoa, but it  
> doesn't work with GNUstep. Now its quite some research work to find  
> out why and fix it.

Sure, but people will help if you want to do it ... however we need  
someone on the SOPE/OGo side, who understands how it all works and  
what the problems are, to list issues in enough detail to make it a  
reasonable way to spend time.  For instance, pointing out what (if  
anything) is wrong with KVC support in GNUstep-base, rather than just  
saying that the area is problematic.

> >  I believe it can also support the extension in libFoundation if  
> > someone asks.
>
> Neither SOPE nor OGo depend on libFoundation. In fact both work  
> just fine on Cocoa.

Because you took the effort to port/develop for cocoa.  I haven't  
spent much time looking at SOPE/OGo, but what time I did spend on it  
was enough to notice some #ifdef options etc to cope with the  
different systems.

> >  But if license is the issue, there is almost nothing people can do.
>
> I'm fine with LGPL, this was never a point for me. I don't know  
> whether it was a strong point for Ovidiu when he started  
> libFoundation, but I don't think so.

I believe Ovidiu objected to assigning copyright to the FSF more than  
licensing.  At least, that was what I understood from my few email  
conversations with him.

> >  As for the place to install (/usr/loca/lib or GNUstep/System/ 
> > Libraries),
> >  gnustep-base works on both ways with right settings.
>
> Even if this is the case (nobody seems to use it!)

I've tried it out, just to check it works ... but of course I'm  
familiar/happy with the normal layout, so I don't do it routinely.   
In practice you are probably correct hat nobody uses it (though I'm  
often surprised by what the silent users of the software turn out to  
be doing).

> producing packages which install it properly also takes some more  
> days(/weeks).

Yes, good packaging is very time consuming.

> I certainly can't write gnustep-make packages which install my  
> software into /usr/local out of the box? I probably need to  
> manually move gnustep-base and do all the right settings etc?
>
> As a rough guess I think it would take about 2...4 weeks to get a  
> (deployable) port. Its the typical 90/10 rule that the last 10%  
> take 90% the time. We don't need a prototype showing SOPE/OGo  
> running on gstep-base, but a solid solution meeting basic QA  
> expectations.

How about spending the couple of days to get a prototype, then  
publicise it and let other people find the rough edges and smooth  
them out?

> Actually I do think that gstep-base is slowly improving (adding FHS  
> etc), but I suppose the issue is that the core developers have a  
> different viewpoint on it (ie they don't think that proper FHS  
> support or Unix/Linux integration is crucial etc).

Well, only Nicola works on gnustep-make generally (you are probably  
the second most expert person on gnustep-make after him) and you know  
how busy he is on other things.   If you have tracked the mailing  
lists, you should be aware that all the core developers are in favor  
of having FHS compliant installation available as an option.  So,  
since you are probably the person best placed to do it, and since you  
know we would be happy to accept such a contribution, why not have a  
chat with Nicola (to avoid any clash if you are both working on the  
same thing) and just add it?

> Anyways. If someone wants to do the port, you are very welcome and  
> of course you will get assistence. I still consider it a goal to  
> move to gstep-base, but at the current pace this will take some  
> additional 3 years. Probably we have moved to Mono till then ;->

As I've said, I'm interested in helping/doing this ... but I've got a  
lot of other stuff I want to do too, so I need pointing in the right  
direction to work in small, managable chunks.  I imagine that, if you  
could do make package stuff (ie install everything in the places you  
want it) and point me to specific issues in the base library, I can  
address the base library issues.