Re: ABI Word Fonts




Independent of all the stuff about "we'd like really nice fonts" wining 
and moaning (justified), here are some current reality and things you 
can do.

Note that anti-aliased fonts, while very pleasing on the screen to look
at for short amounts of time, become tiring to work with if you do so
for extended periods.  We need both better anti-aliased font technology,
and better bitmap font technology (not to mention a unified printing
model; X was being developed about the time Postscript was first coming
out: we didn't try to deal with both screen and printer fonts simultaneously).

1) out of the box, for lots of bad reasons, the font path is often broken,
particularly if you have a >17" monitor.  Fix this in your configuration.
Do what is described in the "XFree86 Font Deuglification Mini HOWTO" at
http://www.linuxdoc.org/HOWTO/mini/FDU.html

2) there is a bug in XFree86's handling of outline vs. bitmap fonts, resulting
in the use of outline fonts at times that would better be served by bitmap
fonts.  Hopefully fixed in XFree86 4.0; this will make things work better
(and change the Font Deuglification process).

3) Truetype fonts are the best (current) way to get pleasing scalable 
fonts; you can/should install them (presuming you have them legally; you 
get a set with Windows and/or Office) and put them in the right place 
in your font path. There are also alot available on the Web.

4) The Adobe Type1 rasterizer available right now for XFree86 is not all 
that great, and produces pretty ugly scaled fonts at screen resolution.

5) I recently spurred discussion of the on the XFree86 list: various stuff
should improve with time.

6) and yes, X's font stuff isn't very good: we designed it over 10 years 
ago, and much has been learned since.  Since desktop UNIX went fallow 
for much decade, X needs to catch up (and preferably pass other systems) 
in this area now that it is alive again.

7) there are better patent free algorithms for generating bitmaps from 
outline fonts than even TrueType (which may have patent problems; there 
is a query from the FreeType folks pending with Apple).  These are getting 
discussed on the XFree86 devel list, and there may now be plans to implement 
them (volunteers, anyone?); they are compute intensive (seriously so) 
and will require serious hacking to do, but it looks computationally feasible 
on current hardware.  John Hobby at Bell Labs came up with these nearly 
10 years ago, but on the iron of the day, it probably wasn't feasible; 
nothing like 20-100x faster machines to make things look feasible...

Sergey Babkin <babkin@bellatlantic.net> and Alexander Larsson 
<alex@cendio.se> seem to be most motivated to try to implement these 
algorithms (John Hobby says that the hassle of trying to get his 
implementation out of Bell Labs is probably not worth the large amounts 
of effort): contact them if: a) you are seriously interested, b) able 
to implement non-trivial mathematical algorithms and make them run insanely 
fast (slow implementations won't be usable).  

It is a two step process: a very compute intensive preprocessing
step you apply to generate font files usable by the rasterizer you run
at run time (which is computationally difficult, but much less so
than the preprocessing step, which has to be run only once).

				- Jim Gettys


--
Jim Gettys
Technology and Corporate Development
Compaq Computer Corporation
jg@pa.dec.com



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]