Turns docs/micromate_protocol_reference.md into a build plan for the missing half
of micromate/ -- it is codec-only today with no way to talk to a unit.
Layout mirrors minimateplus/: framing, protocol, client. Transport is REUSED --
minimateplus/transport.py is byte-level and protocol-agnostic, and already
handles the RV50/RV55 habit of emitting RING/CONNECT to a caller. But
minimateplus/framing is deliberately NOT shared: the two framings differ in ways
that look small and are not, and a shared module would accumulate `if series ==`
branches until neither case is readable.
Records the three things that make S3FrameParser useless on Series IV (bare STX,
0xC5/0x03 flags, uniform 10 XX -> XX destuffing with no inner-frame carve-out),
and the traps that have already cost time: the declared length is a uint16 BE and
reading it as a byte under-reads SUB 0x1A by 47x; SUB 0x1C is four bytes longer
on the Thor line so parse forward not backward, or a BD unit reports 577.92 V;
the monitoring flag must be tested non-zero; and the SUB byte itself can arrive
DLE-escaped, so destuff before indexing.
Scope is READ-ONLY, stated with the reasoning: no command has ever been
originated against a unit by this project, and keeping that true through the read
client means the first thing we ever send to a customer's instrument is a
deliberate decision rather than a side effect of a client that grew a method.
Tests are specified to embed frames as hex constants rather than read fixtures,
since bridges/captures/ and tests/fixtures/ are both gitignored -- with a table
of which cases matter and why, and a round-trip assertion against
scratch/mm_frame_parse.py, which is already known good across three sessions at
zero bad checksums.
Steps 1-2 need no hardware.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ru8Lg9HkkYvX9VWWo65SmL
