[Theora-dev] DC prediction question

[Timothy B. Terriberry]

Marco Bagnaresi wrote:
> Can someone tell me why specification differ from reference implementation?
> In specification blocks L,DL,D and DR are used for DC prediction, but in implementation
> L,UL,U and UR are used. This brings to different results ... or not?
> What's the correct version?
> Thanks in advance

In the implementation, the image is stored upside down (for historical,
Windows-related reasons). Thus what is labelled "up" in the code is
really "down".