Meeting tomorrow or Friday

Tim Rowley tor at
Thu, 2 Dec 2004 13:14:31 -0500

On Wed, Dec 01, 2004 at 10:35:59PM -0800, Brendan Eich wrote:
> 1.  Cairo back end status reports for Mac (Core Graphics, a.k.a. 
> Quartz), Windows (GDI+), and Linux.  We won't replace gfx with Cairo by 
> 1.8final, but we should aim to do so by 1.9.  Roc can say more.

An initial incomplete GDI+ backend has made it to cairo CVS.
That still would leave us with the GDI+ distribution problem -
for that purpose we still need a standard GDI backend.
No signs yet of a Quartz backend.

While mozilla could be code ready for cairo by 1.9, I don't think
it will be feasible to switch for performance reasons (see below).

> 3.  SVG status and plans for 1.8 and 1.9

Our feature set has come along nicely, with more stuck in the
review pipe.  We'd like to get it enabled for 1.8 - currently the
roadblocks are the full-module sr and distribution.  I'm working
on a sr subdivision and schedule right now.

Distribution of the backend rendering library is a big problem -
currently the only way forward looks to be cairo.  We're
currently short two tier-1 platforms, and performance on at least
linux leaves much to be desired even for fairly trivial SVG.
Part of this is caused by the current SVG frame/renderer model
assuming a retained mode backend, but oprofile shows much more
time is spent in cairo and company than layout.   I have some
plans to fix this, but am currently scheduled with feature work.

Profiling cairo leads to some odd results - oprofile shows most
of the time going to cairo/libc and relatively little to the X
server (around 4%).  "top" shows a radically different story,
with the X server taking 70-80% cpu time.