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

Re: State of xfstt



An X font server should become the exception (for XFree86 4.x) rather than
the rule.

It is better to have the server rasterize the fonts than a font server, to
provide TT fonts to legacy X applications (pre Render and Xft).

Font servers should be reserved to providing fonts to a network of machines,
and not the default usage anymore.

In the longer term (as Render and Xft deploys), fonts become a client side
item in any case, with sharing for a network of machines via normal
file sharing protocols (pick your favorite).  The observed fact over 15 years
is that serious applications *ALWAYS* need information buried in the font
files and not exposed in X's (*broken*) font mechanisms; Xft owns up to
this and stops getting in the way of what applications really need.

With 20-20 hindsight, it has become clear that X's font model was wrong,
and that it also hurts performance (it turns out to be a bigger hit to
rasterize a large font to get metrics and transport the large metrics than
to move the (typically small subset) of glyphs from the client to the server.
And font formats have evolved faster than there is any way to evolve a network
of many machines running various versions of X servers: ergo the Render
and Xft approach which reduces X to rendering and caching glyphs with
the fonts themselves on the clients.  And apps need various sorts of font
data not envisionable in the past.

Live and learn.  In our defense, I'll just say we did ok in 1987, when
X11 was being designed.

I hope this discussion helps give some perspective.
                           - Jim

--
Jim Gettys
Cambridge Research Laboratory
Compaq Computer Corporation
jg@pa.dec.com



Reply to: