I'm up against a very difficult decision w.r.t. xupdf direction before
the HT03 article.
The question is whether to
1) devote significant time and effort into making the handling of the large
images practical at the necessary speeds; caching, lazy loading etc.
- we need this anyway for photographs etc, but not quite
yet
- this is rather tricky, forcing to spend time tuning
memory management and caching algorithms
2) devote significant time and effort into rendering the PDF (or possibly
some version made easier through hacking xpdf and writing semi-parsed
output) by direct OpenGL commands, i.e. "render this textured rectangle
here" for a single letter.
- this might lower the performance or not; hard to say yet
- might be able to make good use of vertex shaders
- we want text rendering to be fast anyway
- might be quite tricky to get the images and fonts right;
OTOH might be easier than I think
As you can see, both have large pros and cons. So I'd like to ask for
your opinions. The hardest is that I'm not at all sure which one
will need greater effort and which one will be more useful.