sec08: RSL sweep corrected to canonical Нойманн (dossier rule); CONTEXT note

This commit is contained in:
Dmitry Kokorin 2026-09-26 16:52:48 +03:00
parent 5a34996591
commit 577570c468
2 changed files with 291 additions and 0 deletions

View file

@ -214,3 +214,11 @@
can crawl at ~10 KB/s. 24 Princeton 2015 (643MB) abandoned in favor of Bollingen 1991.
- Monitor-kill trap: pkill -f on the worker pattern also kills the background queue bash
(its cmdline contains the worker command). Kill by comm (python3) only.
## 2026-09-26: sec08 sweep mistake — canonical RU name violation
RSL batch sweep for sec08 used «Нейманн Эрих» (a fallback variant) instead of the
dossier-canonical «Нойманн Эрих». The rule (AGENTS.md Phase 1) was already in place —
I simply didn't check existing dossiers before generating the query list. Fix: before
any batch author sweep, grep every existing dossier for `**RU name` and use the
canonical spelling as the PRIMARY query (variants only as follow-up).