Re: LLVM

[Helge Hess]

Hi,

actually I asked Nicola about LLVM on FOSDEM ;-) License issues aside,  
I also think that it sounds very promising.


First, I dislike this perspective:
---snip---
but this is not a convincing alternative - the license seems designed  
to abuse contributors.
---snap---
This is kinda ridiculous since the whole clang thing was provided by  
the LLVM 'sponsors' (aka Apple ;-). And if Apple releases the ObjC  
frontend too (it looks like they do), it would be quite awesome.
And honestly, how much did we 'contribute' to GCC ObjC in the last few  
years? Nothing, really!
And its not that the GCC people actually like us ;-)


So if we can produce a new and better GNUstep compiler on top of LLVM,  
why not? As long as we are free to use it, I don't really care whether  
Apple grabs changes (couldn't we also put restrictions on that,  
BTW?!). And honestly, I don't think we would provide that many  
interesting things ... It would be most certainly the other way  
around, we would grab the Apple ObjC work and just reuse it for our  
purposes.
I think the important thing is that Apple can't take us away the  
license. If they fork again, their choice.

I guess it would be very much like WebKit. Apple releases the whole  
thing and plenty of FOSS projects benefit a lot from it. Sure, they'll  
still have their own little secrets kept back, but this doesn't lower  
the overall value for the community.


Finally: I do not think that we can get away with it that easily. As  
everyone knows there is Objective-C 2.0, and GNUstep do esnot support  
it. Which is a major reason why I personally slowed down my  
'investment' into ObjC. (an environment separate to Cocoa is pretty  
much like using the 'objc' compiler from David Stes ;-)
I have no idea when the GNU runtime and GNU GCC will support a  
compatible alternative. ObjC 2.0 has been around for 12 months? and no  
one worked on it. Hence I have very little hopes on this :-(
(I talked a bit about GC with Richard at FOSDEM, but we'll see ...)

Greets,
  Helge

PS: the most convincing argument against LLVM I've heard so far is the  
target platform issue. Thats a pretty serious one.

--
Helge Hess
http://www.helgehess.eu/