Re: XFree86 modularization

On the specific issue of FreeType (which is the one I feel somewhat
entitled to comment on), I would tend to agree with Jim.  The
inclusion of FreeType in XFree86 is a bad idea, and causes more
trouble than it's worth.

Had FreeType not been distributed with XFree86, Torrey and his crowd
would probably been more motivated with getting MacOS X-specific fixes
to FreeType getting into FreeType CVS.

Unfortunately, FreeType cannot be removed from the tree as long as
we're using the ``platform-independent'' model for the core fonts
renderer.  Which is something I hope to get fixed in the near future.

The long term solution, of course, is to move all fonts rendering to
the client side.

AH> I know Juliusz monitors the CVS quite closely, I'd have just
AH> thought he'd of shouted fairly loudly if this was a problem.

I agree, I did too little testing for 4.3.

(We did fix all the issues I was aware of at release time, but I
received very little feedback -- mostly from Mike Fabian --, and
didn't do exhaustive testing myself.)

AH> Did the Freetype team make any noise about the bugs that were to be
AH> shipped in XFree86 4.3.0 ? 

They weren't actually bugs, but lack of workarounds for buggy fonts.

FreeType includes a large number of workarounds for fonts that are
buggy according to the TrueType specs, but happen to work with Windows
and hence get shipped anyway.  These workarounds obviously get
developed in a somewhat empirical manner.

4.2 used FreeType 1 for core fonts, which had been tested for a very
long time.  4.3 was the first version to use FreeType 2 for all
rendering.

At the time 4.3 was released, FreeType 2.1.3 was ready, which was much
better than the 2.1.1 that was included.  While Keith Packard did
volunteer to upgrade the included version, it was decided to enforce
the freeze, which, with 100% hindsight, turned out to be a mistake.

                                        Juliusz