References — bi-03 HTML Parsing

Primary sources first. In-tree Chromium docs beat any external source, including these modules.

Specification — read this one properly

HTML Standard §13 — Parsing HTML documents — the normative algorithm. The tokenizer's 76 states and the insertion modes are enumerated here and nowhere else.
Tree construction — including foster parenting and the adoption agency algorithm.
Parsing HTML fragments — a genuinely different algorithm, which is why innerHTML behaves differently.
document.write() — the reason the parser cannot simply run ahead.

Source

third_party/blink/renderer/core/html/parser/ — the whole subsystem
html_tokenizer.cc — the state machine, one case per spec state
html_tree_builder.cc — insertion modes
html_construction_site.cc — foster parenting lives here
html_preload_scanner.cc — speculative parsing and what it may not do
html_document_parser.cc — yielding, and the script-blocking interaction

Conformance

html5lib-tests — the shared parser test suite. tree-construction/*.dat is the fastest way to find pathological cases.
WPT html/syntax

Background

Simon Pieters — Idiosyncrasies of the HTML parser — book-length, free, and the best secondary source in existence.
Encoding Standard — the encoding sniffing that happens before parsing can start.

bi-04 — DOM internals — what the parser hands downstream
bi-13 — Tests & WPT — running html5lib-tests properly
fe-04 — HTML as a platform (sibling project) — the application-level consequences of the same semantics

When a claim here proves stale, record it in PROGRESS.md section 7.