This document outlines issues found with the flif16 specification during implementation. These issues may exist either in the spec doc or in the reference implementation.
- reference implementation ignores loop counter
- spec and reference are unclear of usage of bits per pixel in second header
- The implementation uses the max of those values while the spec does not mention that constraint
- furthermore, the implementation for some reason uses 255 and 65535 instead of 8 and 16 and just takes the log2 of them later (is this some weird optimization?)
- range encoder is not documented
- will there ever be a need for other configs than the 24bit one currently implemented?
- custom bitchances are not implemented and therefore not documented very well