Most compression formats treat the process as a black box: you feed bytes in, you get bytes out. I wanted something programmable and composable, where the format itself can be adapted to different domains — and even customized by different vendors.
So far, CCCP has a few interesting properties:
Composable: Multiple LUTs (look-up tables) and encoding phases can be combined.
Context-aware: Decoding is guided by explicit metadata, not just raw byte streams.
Round-trippable IR: The intermediate representation can reconstruct the original logic before final binary compression.
Programmable: Vendors can plug in their own LUTs, encoders, and decoders.
It is still very early and experimental. Would love to hear if anyone has seen similar approaches, or where this might break down in real-world usage.
Repos: