Criteria for an imageviewer

[Maksim Orlovich]

> SPEED:
> Load images blazingly fast -- no matter how large they are. If they are
> very
> large the imageviewer shouldn't load the whole image but only the amount
> of
> data which is necessary to display the actual view of the image if
> possible.
>
> If you want a rather extreme testcase use this as a reference:
> http://photojournal.jpl.nasa.gov/tiff/PIA03377.tif


That's pretty much infeasible, effort-wise to do that well with all
formats, IMHO (particularly since unlike tiff, the important ones are
actually decently compressed, and I don't think libPNG or libJPEG can do
this sort of manipulation), though libJPEG can subsample [khtml in recent
KDE3 takes advantage of that as a form of DoS protection]. Or are you
suggesting doing seeking over http !?

On the other hand, the decoder lib that's used in khtml (disclaimer: done
by me) in work/kde4 can swap out image data to disk, so it never runs out
of memory, but it might run out of disk -- hmm, need a policy against
that. It doesn't have a tiff decoder, but I've opened 16,000 x 16,000 NASA
images in it (JPEG's, pretty Mars photos). And it does have stuff like
zooming of progressive images, etc. [On the other hand, it's probably
overcomplex for 99% of uses].