Propose the bulk helpers, and a voice pass on the paper

[steve-downey]

Bulk helpers

decode_to / encode_to / decode_into / encode_into already existed in detail/bulk_transcode.hpp, and detail/ is only a directory convention in this repo — the umbrella header includes it and the module exports it, the same as null_term and transcode_string. So they were public API while the paper, the README, and the header comment all said they were not proposed. The tests had been rewritten to exercise the pipelines instead (// decode(codec) | ranges::to — replaces decode_to(codec, range)), leaving all eight helpers with no coverage.

This settles it the other way and proposes them. They are trivial wrappers, and that is the point: converting a buffer to a container is the operation users come for, and withholding the name does not stop anyone wanting it — it only moves the question to every code review and every committee discussion of the paper.

• Paper — the section now runs the concession instead of the assertion: the principled reason not to propose them, then However, we propose them. Overview, user-facing examples, the API comparison table, and footnote 2 follow.
• README — drops "no dedicated bulk API is needed" and "not proposed for standardization"; the four bulk rows name the helpers.
• Header comment — no longer claims they are not proposed.
• constexpr — added to the WHATWG decode_to/encode_to overloads, which lacked it while the pluggable overloads had it.
• Tests — 18 new cases covering all eight helpers, four of them consteval via constify/STATIC_REQUIRE. Each container-returning case is checked both against a literal and against the pipeline it is specified as.

Voice pass on the paper

Separately, a pass over the paper's prose. The abstract opens on the thesis with no wind-up, Context and Motivation is broken into short paragraphs, and the two design decisions that had a real objection worth raising — rejecting char8_t, and the compile-time codec parameter — now argue by concession rather than asserting and then supplying reasons.

Five <!-- TODO(sdowney) --> markers are left in for prose only I can write:

1. The answer to Jan Schultke's epigraph, which the paper quotes and then never addresses.
2. The two motivation sections with the comparison table wedged between them.
3. Prior Art: four paragraphs on the same template, no owned judgment anywhere.
4. The ending, which currently stops on a table.
5. The codec count — the enum lists 40, the prose says 39 in two places.

Verification

make compile clean, 729/729 ctest pass, make lint clean except gitleaks erroring on its own CLI (unknown command "git" for "gitleaks"), which is pre-existing and not a finding.

Not verified locally: the paper does not render here — make transcode-view.html fails fetching wg21.link behind a TLS-intercepting proxy — so fence, div, and table balance were checked by hand. CI should render it. That is part of why this PR exists.

🤖 Generated with Claude Code