OK, I found a legitimate bug as well as generally poor tuning in the short block triggering code. I'm still wondering if the algorithm can't be substatially improved, but at least it's functioning pretty well at the moment. The test sample set (including all the pathologically bad samples folks produced over the last week) now encode entirely correctly. Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/