x.org is Hacker Trash

[Joseph Parmelee]

On Thu, 29 Mar 2007, xorg sucks wrote:

> I had spent so much of my life on this TRAIN WRECK of
> a project that it was virtually painless to make an
> account specifically for this email.

Follows a rant, more than a bit rude, but let's not use that as an excuse to
avoid noticing the many valid points.

> (1) Make an easily navigable site with consistent
> build instructions.
> 
> (2) Concentrate on one method of distribution.
> 
> (3) Only offer individual tarballs for those who know
> what they're doing.
> 
> (4) Setup (easily findable) branches in your
> repository; don't even allow changes to the unstable
> repository unless the whole thing compiles.
> 
> (5) Make it easy. Good Lord! I can compile a frickin'
> OS Kernel without a sweat. What the hell is X Windows?
>

The linux kernel project, arguably the gold standard of open source
projects, puts up release candidates for testing which can be quite easily
built by persons with only a modicum of experience.  Why is that important?
Because it permits widescale testing of all the various different hardware
out there.

There's more to a display system than just a video chip.  No one person or
group, regardless of their size and degree of diligence, can possibly test
their code on the vast number of different hardware combinations out there
without help, particularly when the problem is complicated by ill-documented
and possibly buggy chips, monitors, adapter cards, mother boards, and
BIOS'es.  It is essential that the release candidates get the widest
possible distribution for exactly this reason.  Not only will the X project
be benefitted, but so will the hardware manufacturers as their bugs/features
will more quickly be discovered and documented.

I have been maintaining a private distribution for the past five years,
using literally hundreds of different upstream build systems and code
repositories.  I have to say that Xorg has the absolute worst of all open
source projects with which I am acquainted, by a clear margin.  It makes
mozilla look positively enlightened.  At least they have a tarball and a
build script (even if they insist on mis-documenting its configure options).

I'm not arguing with the value of modularization from a developers'
standpoint, but PLEASE devote the appropriate resources to putting together
usable releases with a master script that allows the unwashed to build this
mess.  There is a vast sea of highly experienced developers out there who
don't happen to be X experts.  They are exactly the people who could most
help you, but they don't have time to play adventure; they have their own
projects to tend.  If they don't build your release candidate because they
can't spare the time to figure it out, you lose the significant benefit of
tests they could run on their hardware, and the help they could give you in
finding and patching the bugs.

> Do it better.

Wikis, README's, and build instruction pages are all very nice and useful
things, assuming they are being maintained, but a tarball with a working
configure is what is most needed.


Joseph Parmelee