>Isaac Yonemoto <ityonemo@xxxxxxxxxxxx> wrote: > >>Why don't we design a new API. We should also do 3, and provide a "quick >>wrapper" for the old binaries. A lot of the encoders/decoders are pretty >>bad anyway, and apparently, the Media kit was broken. This way we can not >>worry about making the Media kit *work* first, and a fluid design paradigm >>will work, too. > >It's almost impossible to provide a "wrapper" for the BeOS R5 media kit >extractor, writer, encode & decoder without having the original header >files. There are not that many anyway, and rewriting better ones is >possible. However, we need an API first. >We can design a new internal API for them, using the information >found when analysing the binaries. We are not limited to the functions >and parameter lists of the current internal functions, but will use them >as a first guidline for implementation. This does not prevent a complete >redesign. >The documented media kit API, as used by applications, will stay >binary compatible with R5. My opinion, for what it is worth, is that undocumented APIs are questionable, at best. I guess if we had some compelling reason to exactly recreate them (say, Productive wouldn't work or something), then maybe. Otherwise, forget it.