R07:L5 (AB half): CipWriter rents its working buffer from ArrayPool and returns
it in ToArray (which now completes the writer; later writes throw). A single-reply
Logix read is one copy of the data (no List<byte> + ToArray); a fragmented read
reassembles in a pooled buffer. EncapsulationClient reads the frame header into a
per-session buffer (reads only happen under the session gate) instead of a new
array per reply. Measured per op (scratch A/B, Release): frame encode 784 -> 200 B,
Logix read 1994 -> 1655 B, 3-fragment read 11.9 -> 8.2 KB, EIP round trip
2.75 -> 2.15 KB; encode time 41-46 -> 28-39 ms per 500k, other timings within
noise; Ab.Transport.Tests 3 s before and after. Modbus transports are left to a
follow-up (its lane is still in review).
Refs #747
Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
CIP codec (#564): PatchU16 bounds guard; CipReply content equality; CipPath
port/length/ASCII/odd/oversize validation with ArgumentException for malformed
tag names and routes; CipPath.MessageRouter/ConnectionManager are now
ReadOnlySpan<byte>; refused Forward Open surfaces ForwardOpenRefusedException
with structured general/extended status; test CaptureFile uses BinaryPrimitives.
EncapsulationClient (#565): frame write bounded by the exchange timeout, which
faults (never cancels) the connection; finite positive timeouts enforced at
every entry point; gate wait bounded by the timeout; deterministic reorder of
the 300 ms test; new dispose-during-drain, gate-wait and RegisterSession
zero-context tests.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
Joseph Doherty