Right now, due to the weakness of mp3 stream format, seeking in mp3 files can be REALLY slow (try to seek in typical 50-100MB file with indian raga...) due to the necessity of generating an index for the file.
What about saving the generated index in a file, for future use?
This could vastly improve seek times for already heard tracks. What's more, the index files could be pre-generated for selected files to allow for fast seeks in the future.
