Subject: X.Org and portability
To: None <>
From: Matthew Mondor <>
List: tech-x11
Date: 07/28/2004 04:48:20
Hi all,

We have noticed lately that the X.Org project is gaining alot of
popularity, to a point where the original XFree86 project is nearly to
a point where it will become much less active in the future. Some BSD
and Linux systems are now migrating to it (notably FreeBSD). The
license related issues are minimal, and do not seem to consist of the
actual basis of the fork. Anyways, the addition of an advertizing
clause making the license "not GPL compatible" is probably the least
of BSD users concerns.

At the current stage, X.Org is very close to the XF86 4 tree still.
There are several concerns which NetBSD users have having pertaining
to it's portability, and the fact that several architectures are still
using XF86 3.3, and they are no less important than other
architectures to us NetBSD users. Moreover, it appears that a fair
number of architectures are still using the older tree, which means
that the XFree86 4 tree would probably require a serious rehaul
towards portability (and so does X.Org currently).

Since I myself have concerns, and am not well versed in the XFree86
internals, and don't have access to those architectures still using
the older 3.3 tree, I was wondering the following if there exists a
definite resource about problems which prevented porting the XF4 tree
to be usable on those architectures? What were the main stumbling

There are plans to make X.Org use an ELF loader rather than the
current XF86 one, which probably consists of a positive step.
Currently, we are using two XF86 trees, which means that there already
is more maintenance involved than what would be necessary for a
single, more portable X11R6 implementation. Moreover, the X.Org
project does not seem to want to drift away from the X11R6 standard or
current XLib API (although there is goal to eventually have XCB
replace XLib (I am not currently aware of what this actually is yet).
Does this mean that it would not be more work to use two parallel
implementations like we currently are using, X.Org for architectures
currently using XF86 4 and XF86 3.3 for other architectures still?
Although in an ideal world, the system should be more portable and
work on all of our architectures with X11R6 support.

Probably that if several people using those architectures using XF86
3.3, and several people using NetBSD in general jump in the X.Org
development team as it is now starting, it would help to make it
evolve portably, at least. Several people are concerned that it would
eventually become i386 centric. The commit rights are currently also
easier to obtain than it was for the XF86 project, apparently. If the
X.Org three is not acceptable to use currently in any standard
distribution, it probably could be supplied in the form of a package
during the portability rehaul work, also?

Just food for thought and sharing my concerns...