Re: Discussing issues

On Tue, Apr 15, 2003 at 04:18:44PM -0400, David Dawes wrote: 
> Thanks for elaborating.  I think I have a clearer picture of what you're
> getting at.
> 
> What you haven't really addressed is why the structure of XFree86 needs
> to change for *you* to achieve your vision.  Follow up your vision, do
> the work, and let it stand on its merits against competing solutions
> (should they exist).  There is nothing about XFree86 that is stopping
> *you* from doing that.

First, I hope you aren't thinking I'm not willing to do work and
pursue my vision on my own. http://www.freedesktop.org has existed for
over 3 years. I've been working on the UNIX GUI for about 6 years, for
5 of those years spending 60+ hour weeks pretty much every week. I
maintain numerous modules and wrote giant hunks of GTK+ and
GNOME. I've written an entire programming book and posted it on the
Internet for free. I've been a Debian developer, and I follow KDE,
XFree86, and several other projects actively even though I don't
typically write code for them.

So yes, I do have my vision out there. But sometimes people working
together can do more than people working alone. That's why I'm
interested in bringing together freedesktop.org, X.org, and XFree86.
That would be a development engine that would be much more efficient,
and effective, than any of them are by themselves.

Much of this current blowup is precisely because there was a
communication gap between these three projects. All the more reason to
bring them together. Agreeing to disagree is easy but ultimately
interferes with getting the best results.

> It might be old-fashioned of me, but I subscribe to the view that
> competing solutions lead to better solutions.  Basically, if you want
> your solutions to become accepted as "the" solution, then *you* have to
> convince those who are going to use it that it is the solution they
> should be using.  Having it somehow automatically "blessed" by creating
> it within XFree86 is not the right way to do it.

I'm not sure you're understanding what I'm saying. Let me try
again. ;-)

KDE and GNOME create coherent, unified platforms by building a single
scalable community that works on a large collection of technology,
with a large number of people involved. In so doing, they've been able
to make much more rapid progress than a lot of isolated sub-projects
on the net would have been able to.

On the down side, because there's so much advantage to being part of
one of these projects - CVS, web, bugzilla hosting; community of
translators and documenters; community of people to help with
questions; faster adoption by OS vendors; etc., etc.  people are very
reluctant to do things *outside* of the larger umbrellas. 

In short, it's much more efficient to do a module underneath some
umbrella. *Much* more efficient.

Current umbrellas that we have, in order up the chain, are something like:

 Linux kernel, GNU C lib, GNU utilities
      or 
 BSD (all three under one umbrella)
    |
 XFree86
    |
 GNOME or KDE  ; Mozilla  ; OO.org ; WINE ; etc.

What's happening right now is that things get artificially sucked in
to GNOME and KDE because XFree86 isn't sucking stuff in. This distorts
the GUI platform to become GNOME/KDE specific.

Now yes, I can create a layer in between XFree86 and GNOME/KDE.  In
fact we already have a thin layer in freedesktop.org.  But to really
beef it up into a strong umbrella organization it would need to start
sucking in implementation code - things like fontconfig, Xft, STSF,
Xr/Xc, etc. and then also new code along the same lines.

Is that what you are asking me to do, or do I misunderstand?

Would you still ship those libraries with XFree86, or would you use
the versions in the freedesktop.org layer?

Would you consider this a fork of part of XFree86 and be angry about
it?

How does this address the issue of server-side extensions such as
RENDER or XFIXES?

Where would you draw the line between XFree86 and my layer?
What do you consider the scope of XFree86 to be - where does it end
and my layer begin? Which parts of XFree86 would it offend you if we
sucked them in?

> If part of this is a policital issue of needing some "independent"
> body to provide a venue for GNOME and KDE to cooperate, then go form
> one.

I already did that three years ago, with reasonable success so far.
http://www.freedesktop.org is the site. It's recently been moving in a
more implementation-focused direction instead of just specifications.

We can take freedesktop.org and beef it up and try to absorb some more
hunks of implementation. What I'd rather do is create an even larger
umbrella group that contains XFree86 and X.org as well, with the same
project hosting, organization, etc.

The reason to do this is that it makes sense all around. It maps the
organization to big-picture technical modules, and lumps together
things that need to be developed as a coherent whole. For example,
does Xft on client side make sense without considering RENDER on the
server side? Can something like RENDER be developed without consulting
the hardware/driver experts?

> If it's an issue of blessed standards, do as Jim already suggested, and
> shop around for the standards body (X.Org, FSG, or one you create
> yourself) that will bless the ones you're interested in.

But I thought you and I agreed earlier that consortiums are broken and
that standards should flow from the implementation project.  In any
case, that is my opinion, even if I misunderstood yours.

Remember, I'm not primarily interested in formal standards, other than
as a rubber stamp. If you go back and re-read my mails in the earlier
discussion of X.org, you'll see what I mean by this.  See
e.g. http://www.xfree86.org/pipermail/forum/2003-March/000817.html

My interest in "blessing" is having a community that really includes
most of the parties (desktop, X.org, XFree86) so that you can create a
discussion and implementation within this community, and have a good
chance that you're addressing the problem in a way that meets the
general need. Perhaps a better word than "blessing" is simply
"coordination."

Coordination does work, I've seen it work countless times.

Havoc