[cairo] Scrolling/Copying

[Mike Shaver]

On 6/23/05, Carl Worth <cworth at cworth.org> wrote:
> If there are parts of cairo that today are slower than they should be,
> I don't want people to train themselves into avoiding them. That's why
> I encourage people to code however they see fit so that we can find
> out where the performance problems are.

That's a fine stance, but it makes me wonder how cairo consumers are
to know if/when given behaviours will be made faster, so that they can
decide if they should do the "natural but slow" thing for their
product release, or add some "fast but awkward" stanza until the
natural path is made faster.  I don't do as good a job of tracking
this list as I perhaps could, but I don't have a good handle on when
things like the tesselation improvements will come on line.  (I also
don't really have time to learn nickel, so I can't really analyze what
I understand to be the primary source for the tesselation code and try
to divine the amount of remaining work from that.)

This issue will come back with a vengeance, I suspect, after cairo 1.0
ships, because there will be pressure on cairo-using applications to
use the system cairo installation, for reasons of consistency and
application size.  The natural pattern might well be much faster in
1.2, but if we have to support cairo 1.0 then we're going to need how
to figure out how to make things fast in 1.0.  (As we will have to do
for whatever performance improvements fall off the 1.0 list due to
resource or API-form constraints.)

I applaud the desire to make natural patterns fast enough that they
can also be the recommended-for-performance patterns, but I'm a little
worried about how to deal with the gaps in light of a now-frozen API
and differences in release schedule for cairo and its consumers.

Mike