Opens Part 2 (the standards in detail). Walks the entire SECS-II
encoding from first principles: the mental model (every value is one
Item; a List is a recursive Item), the format-byte arithmetic
(6-bit format code, 2-bit length-byte-count), the 14 format codes,
length bytes 1/2/3 (with the 16 MiB cap), big-endian everywhere,
the difference between byte-count (scalars) and child-count (lists).
Then walks every format with worked hexdumps: empty list, nested
list, ASCII with length-byte boundary crossing, Binary vs Boolean,
U1/U2/U4/U8, signed integers with two's-complement edges, F4 / F8
with NaN / ±Inf / −0.0, JIS-8, C2 Unicode.
Then the codebase mapping: Format enum, Item variant storage layout,
encode_into / decode_at recursion, SML printer/parser, the
identifier-wildcard rule (SEMI allows U1/U2/U4/U8 interchangeably
for ID fields) with the messages_helpers::any_unsigned_first<Out>
helper that closes the leniency contract.
Closes with the well-defined CodecError conditions, what the codec
deliberately doesn't reject (unknown format codes), and pointers to
chapter 31 (codegen) and chapter 11 (HSMS) as the next dependencies
above the codec.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
raphael
338d0b974d