Re: Light-weight GNUstep distribution

[Graham Lee]

Björn Gohla wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 2003-03-09 22:20, Helge Hess wrote:
>
> > Tobias wrote:
> >
> > > > Here's something I've been wondering.  Has anyone written an
> > > > Objective-C preprocessor and runtime that's compatible with GNUstep,
> > > > but which generates ANSI C code that can be built in (say) Visual
> > > > Studio or another ANSI C compiler?
> > >
> > > there is poc [1].
> >
> > POC is not compatible with GNUstep.
>
>
> but seems to work at least with the watcom compiler under windows.
"Work" can be a little strong a term to use when discussing the POC. 
From discussions with the author on Usenet, the following are not 
implemented (some may have been fixed, though some are deliberately broken):
Multiple root classes.  So no NSObject for us "juvenile idiotic Apple 
developers" (yes, that seems to include GNUStep).
id.  Well, id does work, but the last I saw it was #defined to be void*, 
which has problems if you want to use C code in your ObjC (I certainly 
do ;-) and also seems (according to commentary, though I haven't 
verified this myself) to further reduce the ease with which non-Object 
root classes may be emlpoyed.
Protocols, as previously mentioned this is also so that POC may not work 
with OpenStep-style code.
The message forwarding mechanism is completely different; while this 
would not affect your bog standard ObjC code /per se/ it may 
significantly change the way that a compiled program behaves.  The POC 
runtime is slower than the GNU and Apple runtimes at this.

Also it might be noted that Stes was invited to contribute to the FSF's 
GCC project a couple of years back, and AFAIK has never replied.  They 
wanted to use Objective-C++ and blocks in the mainstream GNU version of 
Objective-C.  He didn't want his code to be used because Apple (or NeXT, 
it may have been that long ago) were contributing to a *different* part 
of the GCC source.  Speaking as an independent observer (I haven't used 
GNUstep/NeXTstep/Cocoa/etc enough to call myself a Stepper, most of my 
code is in plani ol' GNU runtime Object-derived ObjC) who often tries to 
squeeze information out of comp.lang.objective-c, this man seems to have 
a pathological hatred of any Objective-C written after NeXTstep 3.3.
> > > i have not tested it, but i (think i) remember some usegroup speaking
> > > about it working with gnustep.
> >
> > No. The author made clear several times that he won't add the features
> > required for compiling GNUstep code (eg protocols).
>
>
> i could not find what license it is distributed under, but apart from that i 
> would say go for it and implement the features missing to compile gnustep.
It's LGPL as far as I can recall, I remember at some point thinking 
seriously about mending some of the bugs raised above (I've never 
written a compiler, it could be an interesting project).  However, I 
fear that in light of the points raised above, if any attempt *were* 
made to patch the POC to the point where it's useful to 
GNUstep/GCC/etc., it's likely that DS would make future versions of 
(what would then be) his branch of POC closed source.  He would possibly 
also use his position as FAQ maintainer[1] of c.l.o-c to spread FUD 
about how FSF/GNU/Apple deliberately stole his code so that they could 
continue perverting the Objective-C language.  You, he, I would know 
this to be false but FAQ maintainers tend to carry a lot of sway 
especially among newbies.

Right, that's my little rant over.  As I have previously stated, I'm not 
an anti-POC zealot; I've observed the way that David tends to act on 
Usenet and have tried to hold discussion with him on c.l.o-c, 
occasionally even coming down on his side in a flamewar.  However I 
think that at the moment, use of POC in a GNUstep project may not be 
beneficial.  While the code certainly has merit and could be 
(technically and legally) adapted to suit the project's needs, the 
inevitable fallout from David's equally inevitable tantrum would be 
unbearable for a very long time.  That's my GBP1E-2.

Graham.

[1]Which carries an equal quantity of Time Machine Syndrome; I e-mailed 
him nearly a year ago pointing out how some hrefs, software versions etc 
were out of date, and gave suggestions for new inclusions etc.  Still no 
reply and the FAQ hasn't been updated since 2000.
--
Graham Lee, Wadham College, OX1 3PN