On the TI-68k platform, Pucrunch has been used for many years in TIGCC and GCC4TI for compressed programs. The compression ratio is OK, and we have well-optimized decompressors made over the years by Greg Dietsche, myself, and later Samuel Stearley (who created the small/slow version and the fast/larger version).
Pucrunch has proven itself to be an excellent tradeoff between compression ratio and decompression speed. LZMA / XZ yields a better compression rate, but the C decompressor is huge and extremely slow (read: ~1 KB/s).
As FloppusMaximus wrote, it's an asymmetrical compression scheme, which won't work for on-calc compression on a TI-Z80 or a TI-68k, because compression requires way too much memory for those platforms. It could work on the Nspire, though.
