Re: communication

On Sat, 2003-03-22 at 20:03, David Dawes wrote:
> On Fri, Mar 21, 2003 at 07:07:06PM -0500, Mark Vojkovich wrote:
> >On Fri, 21 Mar 2003, Keith Packard wrote:

> >> I have made limited progress in my attempts to address these issues. The
> >> opening of the XFree86 CVS repository and developer mailing lists to
> >> public access has benefitted X development significantly. In order to
> >> improve the technical coordination with other open source projects, I
> >> have become personally involved in many of them.
> >
> >  Yes, we know you have made progress in many aspects of the XFree86
> >project.  Your contributions to this project are why you were a
> >core XFree86 member.  You make it sound as though, despite being a core 
> >member of XFree86, you did not represent the XFree86 project in those
> >actions.  I always thought you were the XFree86 presence in those projects.
> 
> That was my understanding too.
> 
> >You make it sound more like you were merely the Keith Packard
> >presence in those projects.
> 
> That's unfortunate.  I've been wondering why Keith has been bringing
> up the lack of communication between XFree86 and those projects as
> an issue when we all thought that *he* was the key part of that
> communication channel.  It sounds like the best thing for the future
> would be for more representatives from those projects to participate
> directly in XFree86 development discussions ([email protected]).

I think perhaps the term "communication" here is not really representing
the real problem. The real problem we've seen is perhaps one of 
interest.

In general, beyond Keith and Jim (and to some extent Mark), I've seen
very little interest among XFree86 developers in:

 - Updating the X rendering model to match current expectations

 - Specific needs of applications not being met by X currently.

 - Application performance (not how many rectangles can be drawn
   per second, but how the things apps are actually doing perform) 

 - Protocol issues in general
 
As evidence for this, I'll put forward:

 - Lack of participation on [email protected]. (People complain
   about render not being suited to hardware implementation, but
   discussions about it were public and advertised to XFree86
   developers from the very beginning.)

 - Lack of participation on wm-spec-list and xdg-list.

 - Lack of participation on desktop or toolkit mailing lists.
   (Though maybe everybody is just in the KDE camp... since
   I don't monitor those mailing lists.)

 - Lack of followups to mails from me to [email protected] on various
   of these issues. (While XFIXES may not have been handled very
   well, it's not like the issues that it tried to address weren't
   brought up long before without any interest or comments.)

Now you might say that the current XFree86 people were too busy 
with hardware drivers and general bug fixage to worry about this
sort of stuff, and that nothing is keeping other people from
getting involved.

But the fact that nobody other than Keith was working on these
issues and getting stuff into XFree86 makes one wonder about structural 
problems; from my perhaps biased perspective, this stuff is really
a lot more interesting than support for flavor-of-the-day of some
chipset, so you'd think that interested developers would be 
crawling out of the woodwork. Perhaps the problem is that
without current developers around to discuss, and review changes
in this area, there is no way for developers to become engaged
and eventually trusted.

Anyways, I just wanted to give some more details on where I see the
problem; it's not necessarily that people are shutting their ears
to the needs of the desktop, it's just that the needs of the
desktop (other than for a working X server) don't seem an 
interesting problem for the bulk of XFree86 developers. 

The desktop projects don't just need the assurance that they
can mail [email protected], they need the assurance that there
will be someone there paying attention.

Regards,
                                          Owen

P.S. - the above contains a lot of sweeping generalizations. I'm
       sure there are lots of places where you could show
       isolated counterexamples, but I don't think they disprove
       my points.

P.P.S - how does this all relate to the issue of governance? 
        I won't claim there is an exact connection, but I would
        make two points:

        The current system where direction is by the consensus 
        of people working on XFree86 for a long time is, I think, 
        very vulnerable to the project becoming isolated to a narrow
        set of concerns without any realization of that or intention
        in that area. Clearly having someone forcing developers 
        to work on stuff they aren't interested in working on
        (or can't work on because they are paid to do drivers)
        isn't going to work, but I think there would be major
        value in having some group able to look at the big picture
        and try to figure out how to get people involved to address
        areas not being addressed.

        Second, I think XFree86 is discovering something here that
        has been discovered by many other projects previously.
        Direction by consensus of developers works great only
        as long as there is consensus of the developers. When
        consensus breaks down, having a body that is formally
        charged with the smooth running of the project can do
        wonders for getting disputes resolved in a constructive
        way.