
Originally Posted by
fredrp
it's binary data that's mostly a sequence of "messages" that are all of the same form, with some text-like content. No audio, no images. Random access not necessary. 300-1500 bytes, tops. Right now I am getting on average a 40% reduction with ZIP.
I would think that maybe a single dictionary could be developed by surveying a bunch of candidate data streams (which I have), and then NOT storing the dictionary in the compressed stream, but have it stored in the decompressor. I'll take a look at LZSS and BWT. Any pointers to source code for these?