A cracker, and the words that kept getting confused

Four things sound like each other and are not:

	what it is	reversible
hashing	name to eight bytes	no
encoding	those bytes to eleven printable characters	yes
decoding	those characters back to the bytes	yes
cracking	guessing names and hashing them until one matches	the only way back

decode already existed and is the one that misleads - it opens the envelope and leaves you holding the shreds. Its own doc comment says so, which suggests this confusion is older than today.

obscene-tool crack is built (D049). Given a list of NIDs and a list of candidate names it hashes each candidate once and reports the matches, with a header recording the suffix, how many candidates were tried, and how many known pairs the list reproduced.

Self-test: fed the 389 harvested pairs as both targets and candidates, it recovers 389 of 389 and reports the generator as reproducing 389 of 389. The corpus is now data/nid-corpus.txt with its provenance written into the file.

That corpus is worth more than it looks. Every pair in it was produced by somebody else's implementation of the same hash, and all 389 agree with ours - an independent check on the whole chain, obtained from log output that cost nothing.

What is deliberately absent: a candidate generator. Describing the naming convention is the actual problem and it wants iterating on, so it belongs in a script that emits a word list rather than compiled into the tool. The tool does the fast exact part.

---