On Tue, 29 Aug 2000, Stefan Seefeld wrote:
> Cesar Crusius wrote:
>
> > i think what ggi really needs is a deep breath, and a deep cleanup and
> > reorganization. if you download the sources, they are a mess. there
> > are sources that don't compile at all, and are known not to compile.
> > libraries are hidden in a sea of directories, etc. imho the
> > distribution has to be rethinked. as it stands now it kind of scares
> > potential users/developers. here's my take on it:
> >
> > i think it would be better for the future of ggi if a new release
> > started from scratch, in small steps. so you could release first just
> > libggi and a few drivers (X and framebuffer, for example). forget the
> > svga driver and others for a while, and don't even think of XMI and
> > stuff. a small and clean distribution that runs on X and the console
> > is an excellent first exposure to the project and its philosophy. it
> > needs some documentation, but the programmer's guide is not that bad.
> > when the release is ready, post it on comp.linux.something and people
> > will look at it, believe me. just don't make the other stuff available
> > or people can get scared. point: don't show something that doesn't
> > look very good.
>
> I agree on the diagnosis but not on the therapy.
Me too. You have to show the incomplete stuff to potential
hackers or they don't see any place to contribute, at least not without
digging more heavily into the guts of the GGI architecture than most
newbies are willing to. Just be sure to clearly identify what's
'production quality' code a what isn't, and people won't get sandbagged.
> We had a similar situation in spring with berlin, when our source code
> got pretty big due to new language bindings
> and extensive docs. What we
> did was simply to refactor the code into sub packages and release them
> separately. You still can download Berlin-0.x.x.tgz but you can download
> individual packages as well (the server code, the various client bindings,
> the docs etc.). This not only makes packages smaller, it also enables you,
> much as Cesar suggests, to separate self contained and stable code from
> more experimental one. The packages can have independant versions, you can
> even release binary distributions for the stable parts.
> Yes, I think that definitely would help.
Agreed.
Jon
---
'Cloning and the reprogramming of DNA is the first serious step in
becoming one with God.'
- Scientist G. Richard Seed
