Owen Taylor wrote:
On Mon, 2006-11-13 at 08:17 -0500, Adam Jackson wrote:
Jesse Barnes wrote:
Well, I created an account but can't seem to edit the page. Anyway, I
like the idea of getting rid of xfs, it seems fairly useless. And to
quote mharris from awhile back:
Long term, what would be really nice, is if someone figured
out a way to implement core fonts using fontconfig/Xft
underneath so we have one font system, and it provides
legacy compatibility to ancient applications that have not
been updated to modern interfaces. That is something that
would need to be spearheaded at the X.Org level rather than
at a specific distribution however.
So here's where I start to admit some font ignorance, but. I took a
hack at this a while back, and it seemed to me like using fontconfig for
matching was the wrong idea, since the properties available from
fontconfig didn't quite match what you can select on with XLFD.
As said on the wiki page, what problem is xfs trying to solve? If it's
just about the ability to add fonts at runtime without having to do
'xset fp rehash', then that should be a very small amount of code to add
to the X server. You need to watch for directories popping in and out
of existance, but, okay, we know how to do that.
There's basically one reason why we switched to xfs years ago ...
performance with large fonts; if you don't use xfs, then trying to
load a large core font (for CHJK) can block all requests to the server
for several seconds. Well, it could on machines of that day.
Since core fonts are really fringe at this point, the main concern with
a switch back to using in-server fonts would be configuration ... do you
really want to change complicated %post scripts for a bunch of packages
to do something new and different?
Would you even have to though?
It looks like %post for the core subpackages of xorg-x11-fonts is just
doing chkfontpath/mkfontscale/mkfontdir for the core stuff, and then
fc-cache for fontconfig visibility. Okay, fine, you might have added a
directory, great. The server (in the new world order) would just check
directory mtimes at font open time and automatically rehash and so
forth, so you're doing _more_ work than you need to at install time. Right?
In which case, "fixing" font packages is a matter of requiring a
sufficiently new server and deleting %post. I am totally okay with fix
by deletion.
- ajax
--
fedora-devel-list mailing list
fedora-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-devel-list