Xiao & Klynt
Xiao
Klynt, I found an old ARPANET packet trace in the archive—only a handful of packets still show the raw header fields. The structure is so oddly efficient; maybe there's a pattern in the padding we can use for a new compression trick. What do you think?
Klynt
That’s a gold mine for a low‑overhead compressor. The padding is basically 0xFF repeats, used to keep the checksum on a word boundary. If you strip the padding and then run a simple LZ‑style back reference over the remaining bytes, you’ll get a decent ratio without touching the header. Just be careful not to corrupt the checksum field—those old packets were picky about that.
Xiao
Sounds solid—strip the 0xFFs, keep the checksum intact, then LZ back‑references on the payload. Just double‑check the word‑aligned checksum after each modification; I’ll run a test on a sample packet to confirm.
Klynt
Good plan, just make sure the checksum stays a stone. Let me know how the test goes.