Alhexx: I *think* NeutopiaW said the normals are stored in what we thought was the 'small model'.
Cyberman: Yeah, decoding LZSS is fairly easy, since it's just following whatever the coded data says to do. Efficiently compressing it is more of a challenge; I never managed to get anything as quick and efficient as the 'original' LZSS code written by Prof. Haruhiko...
You should only need to buffer 4K of the output IIRC, although the buffer is a good idea for speed.
And yes, the best way to make sure you've got decoding right is to decode a file using your routine and a 'known working' routine (I *think* mine would count as that
) and see if you get the same output...the parts that are hardest to get right (well, not *hard*, but easy to miss) are making sure that repeat runs, and reads from before the beginning of the file, give correct output. My code didn't do repeat runs at first, but I didn't notice for a while...