> I think a library is needed. It should be fairly minimal. Perhaps just > color conversion etc. Why should my application have code written > specifically for every different broken webcam that exists? > > Some kind of extensible minimal library may be desirable. That way we > could get the webcam decompression code out of the kernel. No more > closed source kernel modules... That wouldn't fix the closed source issue, it would just move it from kernel to user space. > I would say a minimal, easy to use, library would be used. I would say > that every v4l programmer started with xawtv and worked from there. I doubt that. I've often seen complains saying that the xawtv code is too complex. So probably many people have a look, but very few actually use it. > If xawtv were re-written to use the new library then things should be > sweet. xawtv already comes with a library, some time ago I started separating code and moving it to the libng subdirectory in the source tarball. color space conversion is there, a abstraction layer for driver access is there (v4l / v4l2 / bktr), and some other stuff. I don't feel like swapping that for something else (which will put you back to square one -- nobody using libv4l). There is very much existing v4l code which doesn't use a library, which will make it very hard to establish a library as standard. It worked for the alsa people because the library was there from day one, and there is absolutely no documentation about the kernel API (thus you are forced to use the library). Gerd -- #include </dev/tty>
