Table of Contents
Provenance: programming
Working log behind programming. Corpus-wide caveats: corpus. This sitting wrote six namespace outlines together; the joint query log, freeze, and review accept/reject table are on design. This page adds no bibliography keys. No ~~DISCUSSION~~.
Run: 2026-08-27. Item cursor-drain-ns-overviews (item 182, run 67). Creating: programming did not exist (export_raw was the HTML “This topic does not exist yet” error page).
What [[programming]] publishes from this sitting
| # | Query | Result |
|---|---|---|
| Q1 | extraction size | 5,859 |
| Q3 | crawled | 1,120 |
| Q9 | venues | 7 |
| Q10 | live children in this namespace | 17 exists, 0 missing |
| Q11 | stub bytes: CrUX / Similarweb. multilingual rebuilt 2026-08-27 | 8,394 / 4,514 |
Re-run: node scripts/report_namespace_overviews.mjs. The script fails if any of those 17 goes missing. Stub-ness is a judgement (short, not corpus-backed), not a byte threshold: the report prints the live byte counts so a later sitting can see they moved, and does not fail on a length change.
The two remaining stubs are the load-bearing judgement on this outline. A reasonable person would have hidden them or written a corpus-backed rebuild. Hiding a page start already links is a lie; rebuilding the instrument notes is a different drain item. programming:multilingual_support was rebuilt 2026-08-27 (this file's Q11 originally counted it as the third stub at 8,012 bytes).
Where the deep figures live
Not here. crawler, openwpm, interaction, traffic_files, tranco, docker, registration, stateful_stateless, cloudflare_radar each have their own report and provenance page. This outline does not copy their headlines.
Review
See review_log.
Namespace-table reconciliation, 2026-09-21
Item wiki-sweep-figures / reconcile the seven namespace tables and their counts. Joint log, full report output, mutation test and the literature: decision: namespace-table_reconciliation_2026-09-21.
This namespace was the one with a wrong count, and it published two of them. The WRAP box said “All 19 children below are written”; Methodology said “The 18 / 2”. The live namespace holds 22 pages. Crawler detection, LLM agents and Deployment had no row in any table here. All three were added — detection in a new When the site notices section, LLM agents under Driving the browser, deployment alongside Docker in a section retitled What you record, how you pin it, and how you run it — and both published counts are now 22.
| # | Query | Result |
|---|---|---|
| Q10 | pages whose id begins programming: (core.listPages, depth 0, provenance: excluded) | 22 |
| Q11 | rows in this page's tables linking inside programming: | 22 before: 19 |
| Q12 | of the 22, short instrument notes not re-derived against the extraction | 2 — CrUX 8,394 bytes, Similarweb 4,514 bytes, unchanged since 2026-08-27 |
The 2 is a judgement about depth, not a measurement, and the page now says so rather than presenting it beside the 22 as if both came from the same check.
Re-run: node scripts/report_namespace_overviews.mjs (exit 0 = every namespace table equals its namespace as a set, both directions). Guard mutation test: sh scripts/report_namespace_overviews_mutations.sh <dir>.