The channel with no line

casino-tax-distribution has carried one open item through four phases. The constitution earmarks 34% of the gross casino revenue tax for a county student fund that reaches every district, and the substitution critique of earmarked gambling revenue asks whether the General Assembly simply set foundation aid lower because the money exists.

The appropriation history was retrieved and did not settle it. In constant FY2025 dollars the enacted foundation aid appropriation moves by a median of $236 million a year, so a substitution smaller than that is arithmetically inside the year’s other reasons for moving. The node then named exactly what would settle it: the casino distribution’s own size, “to establish whether the channel is even of a magnitude the total could register” — and recorded that both publishers, tax.ohio.gov and casinocontrol.ohio.gov, return 404 at their document root to a self-identifying agent.

That last sentence was true and was not a reason. tax.ohio.gov serves its pages to a browser, and the casino page links every document under /static/, which resolves to dam.assets.ohio.gov/raw/upload/tax.ohio.gov/…. That is the route tax-abstract has used since it was wired and dot-sd1-school-district-taxes.md wrote down — for a different division of the same department. The channel was declared unreachable for two phases behind a mapping this repository had already documented.

What is there: eighteen half-yearly per-district distributions, August 2015 through January 2024, in sixteen workbooks. Nine complete state fiscal years. Between 979 and 1,044 districts each, keyed by (county, IRN) because a district is paid out of every county its students live in.

Declare tax-casino as a twenty-first connector, Wired with a recorded remainder, feeding revenue-stream, education-agency and metric; build crates/dispersion/fixtures/casino-district-distributions.csv from it, one row per district per distribution; and read it through dispersion::casino.

A second connector on a publisher already in the registry, for the reason ohio-bills was a third on a publisher already there. tax-abstract retrieves what a district may levy; this retrieves what the state hands it outside any formula. Different division, different directory, different remainder, and folding them together would give one connector two blockers and one status that could describe neither.

The grain is the payment and not the year. Money moves twice a year and the halves are not interchangeable: January–June 2020 came to $24.6m against a $42.9m floor across the other seventeen half-years, and an annual figure averages the closure away.

The open item closes, negatively and conclusively. At $114.2m in FY2024, its largest year, the whole channel is under half the $236m median annual movement in foundation aid. The null result stops being “we could not see it” and becomes “it is not visible at this size, and no longer series will change that.” That is the one thing the record said was missing.

And it lands the node’s own thesis exactly. The lottery’s substitution was legible in a $97.6m movement — smaller than this channel’s largest year — because Fund 7017 is itemized inside the appropriation table. Casino money never enters the table, so the only test is inference from a total whose noise is twice the size of the thing being looked for. Detectability follows from itemization, not from magnitude, and the two channels now demonstrate that at comparable sizes rather than asserting it.

A new remainder, which is a publisher’s silence rather than a gap. Nothing after January 2024 exists on the department’s page in any category, and the distributions before August 2015 are PDFs. still_blocked says both, so a refresh that finds nothing is not read as a failure.

A twenty-first connector also means the approval list grows again. It is derived from these records precisely so that growth is deliberate, and a connector appearing in the registry without one keeps failing the test that first caught it.

Leave issue #12 closed as a null result. Rejected: “no offset is visible” and “no offset could be visible” are different claims, and the node was carrying the first while wanting the second. Without the magnitude the reader cannot tell whether the appropriation series failed to find something or was never capable of finding it.

Fold the sources into tax-abstract. Rejected, above.

Key the fixture on (county, IRN). Rejected: 48,000 rows to carry a fact that a count carries. counties keeps the apportionment visible — 1 for 294 districts and 88 for three e-schools — and makes it assertable, without a panel five times the size in which no calculator is interested.

Carry the revenue half-year as a column. Rejected: it is a statutory consequence of the payment month, and a derived column in a fixture is a place for two things to disagree. It is derived in dispersion::casino and checked at build time against the fifteen sheets that print it — a rule checked eighteen times a rebuild is not an assumption.

Reach for the Internet Archive first, as the record suggested. Unnecessary, and it would have cost the finding that matters most about the end of the series. What establishes that the series stops in January 2024 is the live page carrying nothing later in any category — an archived copy of a page dated earlier could not have said that, and one dated later would still leave the question of what had been removed since.