nx_arxiv_census.nx

nx_arxiv_census.nx -- COMPLETE-POPULATION CENSUS of an arXiv date window. Turns "I read the current month" from a seat's assertion into a MECHANICAL verdict, which is what the sotabar/barscan law needs. COMPOSES the incumbents, deliberately does not duplicate them: se_fetch (nx_search_extern) -- polite sovereign-TLS fetch, the ONE arXiv fetcher em_int_after (nx_search_emerge) -- opensearch:totalResults reader em_collect_yms (nx_search_emerge) -- <published> year-month extractor em_maxym (nx_search_emerge) -- newest YYYY-MM, which IS the barscan newest= field ac_reason/ac_verdict (nx_arxiv_census_lib) -- the ONE decision, shared with the gate in-process The ONE question neither incumbent asks: does opensearch:totalResults RECONCILE against the number of entries actually delivered? Equal means the window is a POPULATION and an absence claim over it is sound; fewer means a SAMPLE wearing a population's clothes and absence over it is UNPROVEN. nx_search_emerge is NOT defective and this organ does not replace it: it samples 12 on purpose for a cheap daily velocity signal and HONESTLY prints "sampled hits" beside "total volume". That is a different question (is this field moving) from this one (what did this window actually contain). Its em_collect_yms cap is silent at the FUNCTION level for any future composer, which is filed as dataio rung DI19. license_tier: ORIGINAL