# Review of the priority queue

The queue is a useful, reproducible intake portfolio. Its first 100 entries contain five editorial selections from each of 20 domains. All 134 editorial seeds are retained; the other 866 records are provisional catalogue matches. The explicit unread, unacquired and per-book-rights-review states are important strengths. Editorial selection expresses a reason to investigate a work, not a completed reading or demonstrated marginal concept yield.

**Describe the author cap precisely.** The code limits the first catalogue attribution to three records after removing its role label. It does not count all coauthors or constrain translators across different source authors. Legge mediates both Laozi and Confucius; translator-only records place Müller or Waley in the primary-key position. An editor listed first can similarly stand in for a collection’s authors. Display “three per first catalogue attribution,” and later inspect author, translator, editor and source-voice diversity separately. The cap is useful spread control, not a measure of representational diversity.

**Treat work families as provisional.** Title/first-attribution normalization removes subtitles, edition annotations and volume/part suffixes. It can merge distinct components while missing alternate titles or editions with changed attribution order. The Sophocles entry itself contains three plays, and some other entries contain selected stories or several works. The defensible count is 1,000 distinct Gutenberg IDs, not 1,000 independently established literary works. Retain edition and component relationships during intake rather than discarding them as duplicates automatically.

**Keep dates and voices separate.** `pgReleaseDate` is a Gutenberg release date. It does not date original composition, translation, selected edition or scientific validity. The first 100 include historical medical, scientific and psychological texts whose value may concern inquiry and changing explanation rather than present guidance. Keller’s contributors, Dickinson’s editors, Shevchenko’s translator/biographer and translated traditions require separate attribution. An English baseline improves practical access while inheriting the archive’s language, period and publication biases.

**Read rank as an acquisition sequence.** Equal 50-per-domain targets and interleaving are editorial choices. They prevent a large category from absorbing the queue, but do not establish equal importance or equal concept novelty. The numerical heuristic and low-ID tie-breaker are scheduling devices. A metadata match can place an unexpected book usefully—or misclassify it. Keep the exclusion log reviewable: a work called a “journal” can be personal testimony rather than a periodical.

The frozen catalogue contains 79,381 records, including 61,856 English Text records; the homepage count 79,386 is a separately observed figure. Preserve both. After each small reading batch, record actual reusable distinctions, overlap with Carrier, edition problems and worthwhile disagreements, then revise later priorities. The 10,000-concept aim should remain separate from per-book quotas.

Reviewed sources: `priority-1000.json`, its first 100 records, the supplied catalogue metadata, and `tools/rank_catalogue.py`. No queue records were changed; this was not a full-text evaluation.