D020 - What the target is, measured rather than looked up
Firmware version decides nearly everything else on this platform: which jailbreak works, which payloads run, whether a game needs backporting. It was the one fact this tool could not answer and everybody had to find elsewhere.
The shell answers sysctl, df and
ps. Each was run against a target and the parsers were
written from what came back, not from a manual page for
a different system with commands of the same name. sysctl
prints a hex dump rather than a value, so a fact is reassembled from the
bytes - not from the dump's own rendering column, which substitutes a
dot for anything unprintable and cannot be un-substituted
afterwards.
Two keys are absent from the list on purpose:
machdep.idle and hw.physmem answered no
such file or directory, so asking for them would show a permanent
gap that looks like a fault.
The measurement corrected the design twice.
df returned 1183 filesystems, which read as a parser bug
and was not: 22 are the machine and 1161 are bind mounts inside running
applications' sandboxes. They are counted and shown behind a fold rather
than discarded - they are real - but listing all 1183 flat would bury
the figures somebody came to read. And ps marks a title
only for processes that have one, so the column is blank for most rows;
finding it by counting across would have put a memory figure where an
identifier goes.
The target test asserts shapes and prints only field lengths. Firmware strings, model numbers and account identifiers are facts about somebody's target, and a test that pinned them would fail on everybody else's.