It is theorized that the Adobe Flash Player renders frames with 4× FSAA (take a
look at a histogram of a rendering). This behaviour should be matched in SWFDec
to avoid rendering errors caused by premature anti-aliasing.

To achieve this:
 • Anti-aliasing must be disabled
 • Frame must be rendered at 4× final dimensions
 • Frame must be resampled to final dimensions with sufficient per-pixel samples
to get good results

I have been hacking around with the ImageSurface, which takes 4 samples when
painting a scaled source surface. Because of this, it's necessary to resample
over several size-halving iterations to maintain the best quality rendering. The
process of creating and painting these intermediate surfaces is far too slow for
real-time rendering, but then I don't think we're using ImageSurface in the
backend anyway - I think it's an XlibSurface provided by GDK.

What I'd like to know is first of all how well SWFDec can render at 4× size,
without performing the resampling. In my experience with Cairo for Python,
things get slow very quickly, and this worries me somewhat. Is the bottleneck
actually in painting the massive surface to screen (which wouldn't be necessary
if we resampled)?

Unfortunately due to broken Ubuntu Edgy repositories, I'm unable to get the
build dependencies for SWFDec right now so I can't answer this question myself ATM.

Thanks