The acts themselves

lsc-budget’s blocker had one clause left standing after two phases of correction: before FY1999 there is nothing, because the greenbook series begins at the 124th General Assembly and the Catalog of Budget Line Items reaches only FY2002. Both floors were probed and both are real.

What issue #7 recorded as the remaining hope was the session laws — the enrolled acts rather than LSC’s analyses of them — and it recorded them as not yet probed, with the right caution attached: “an act’s text and an act’s appropriation table are different things and only one of them answers this.”

They are the same thing. Ohio’s operating budget acts carry the appropriation table in the bill.

Add ohio-session-laws as a connector in its own right rather than folding it into ohio-laws. Same publisher, different artifact: ohio-laws retrieves Revised Code sections as they stand today, and this retrieves acts as they were passed.

Wire the two acts whose education tables are printed once: Am. Sub. H.B. 215 of the 122nd (FY1998-99) and Am. Sub. H.B. 282 of the 123rd (FY2000-01). Reconcile every parsed row against the act’s own fund-group totals and those against its grand total, in both fiscal-year columns, and fail the rebuild on a difference of one dollar.

Do not wire H.B. 650 and H.B. 770, and record why in a form a later phase can act on.

Keep the four years out of the published feed for now. They are in the crates, tested, and joined; the site’s series still begins at FY2002. FY1999 cannot go on a chart until the shape it needs is built, and putting it there without that would draw fifty-one programs falling to zero.

The floor is the publisher’s and it is stated, not inferred. GET https://search-prod.lis.state.oh.us/api/v2/ returns the service’s own index of every General Assembly it holds: seventeen entries, oldest the 122nd, 1997-01-06 to 1998-12-31. Nothing below it exists in any version code. So the enacted record reaches FY1998 and no further, and the Foundation Program era and the equal yield formula are unreachable by this route — not unattempted, unpublished. That is a different kind of blocker from the one this issue carried, and a much better one to have.

DeRolph I was decided in March 1997 and falls inside the 122nd, so the first budget enacted after the decision is the oldest thing there is.

And that act declined to itemize half of its own biennium. H.B. 215 spreads FY1998 across fifty-three GRF education lines and FY1999 across one: fifty-one of the fifty-three are $0 and the whole year — $4,470,135,592 — sits in 200-405 Primary and Secondary Education Funding, under prose promising an itemization by 15 January 1998 and making FY1999’s earmarks “subject to the passage of a new school finance formula and education reform plan”.

The zeros are the act’s, not the reader’s, and the reconciliation is what proves it: 53 GRF rows sum to the printed 4,899,708,534 and 5,134,145,592 exactly, in both columns.

The promise was kept, and the evidence needed no new retrieval. H.B. 650 itemized FY1999 in February 1998. What shows it is a fixture this repository already held: the greenbook’s FY1999 actuals carry 141 line items and no 200405 at all, while 200501 — appropriated nothing — spent $3,035,363,396. A line appropriated $4.47 billion and spent zero, because by the time the money moved it had been given its proper names.

Two defects in this repository’s own published figures, both found by reading the acts. TAX_REIMBURSEMENT excluded 200417 and did not exclude 200906.

• 200906 Tangible Tax Exemption - Education is a reimbursement by the module’s own definition and sits beside 200901 in every table from FY1998. Omitting it counted $73,500,000 inside the department in FY2002 — 0.94% — decaying to $10,707,622 by FY2009. Because it phases out it inflated the early years more than the late ones, which is the same wrong-sign error the module’s docstring records catching once already. FY2002-FY2013 all move.
• 200417 is Professional Development from FY1998 to FY2003 and only becomes Personal Property Tax Replacement Phase Out in FY2024. Excluding it by number was harmless while the series began at FY2002; it stops being harmless the moment FY1998 is read, where H.B. 215 appropriates $14,370,077 to it.

The exclusion is now keyed on the number and the year. And the test that was supposed to guard the class had the same blind spot as the list — it matched titles containing “allocation”, “replacement” or “reimbursement”, and 200906 says exemption.

The 123rd took education out of the operating budget. H.B. 283 is that biennium’s budget, 977 pages, and has no Department of Education section and no 200-5xx line anywhere. Education was appropriated by H.B. 282, enacted a day earlier, as the response to DeRolph II. “The operating budget” is not a reliable way to find the education table.

Three retrieval facts worth having written down. The enrolled version code is the bill’s position in its own version sequence, so it differs per bill — 06_EN for H.B. 215, 05_EN for H.B. 650 and H.B. 770, 08_EN for H.B. 282 — and guessing wrong returns a nine-byte 404 that reads exactly like “not served”. The index at .../legislation/{bill}/ gives the right one. Every act served the project’s own user agent without complaint.

One act’s arithmetic is wrong and is carried rather than absorbed. H.B. 282’s printed GRF total for FY2001 is one dollar more than its own fifty-two rows, while its group totals sum to its grand total exactly. It is listed as a named defect of a fixed size, not swallowed by a tolerance — a tolerance wide enough to hide a dollar is wide enough to hide a missing row.

What the reconciliation caught, which is the argument for having it. H.B. 215 closes its lottery funds with Total 017 and 018 LPE Lottery Profits Education Fund Group — lowercase, named by fund number, and unlike every other total in the act. A reader matching TOTAL misses it and loses $699,892,200 silently. The check failed instead.

Wire H.B. 650 and H.B. 770 too, and publish FY1999 itemized. Rejected for this phase and scoped rather than abandoned. Both print every amended row twice — struck inline, inserted beneath — and H.B. 650 reprints H.B. 215’s fund-group totals unchanged, so the reconciliation that guards every other extraction here would pass against the superseded number. That is the worst possible property in a source: a wrong answer that satisfies the check. Reading them needs a reader that tells a struck row from an inserted one, and it needs its own phase.

H.B. 650 also carries a second education table that was never law — SECTION 31, $15,000,000 from a School Trust Fund, conditional on a ballot measure that failed on 5 May 1998, with line items printed as literal placeholders XXX 200-YYY. Whoever wires H.B. 650 must exclude it deliberately.

Put FY1998-FY2001 in the feed now. Rejected. FY1999’s row set is fifty-one zeros and one lump; charted, it says the state stopped funding transportation, special education and vocational education for a year. Carrying it needs a per-row itemised/lump distinction and a vintage field naming which act a figure is read from, and both are design decisions worth making deliberately rather than at the end of an extraction phase.

Treat the FY1999 lump as a data defect and interpolate it. Never on the table. It is the most interesting thing in the four years.

Take the totals only and skip the line items. Rejected. The line items are what make FY1999 legible: a total for FY1999 looks ordinary, and it is the shape of the table that carries the finding.

Look for the pre-1997 acts somewhere other than the legislature. Attempted 2026-08-27, and the answer is better than the negative it was expecting: the acts are not lost, they are licensed.

HathiTrust holds the bound Laws of Ohio for 1853-1921. The Internet Archive’s fifty-nine title matches are all 1881-1942 topical compilations and not one is a session-law volume. HeinOnline holds 1803-2022 complete, by subscription. LSC’s own searchable record begins 1997/98 — the same Legislative Information Systems floor this decision establishes from the publisher’s index, reached independently and agreeing with it.

So the era this decision cannot price is digitized and behind a paywall rather than absent. The span that matters most — the Foundation Program era, the equal yield formula, and the years DeRolph was tried on — falls in the gap between HathiTrust’s 1921 and the 1997 floor, and is served by exactly one publisher, commercially.

That changes the ask rather than clearing it. It is a request to a license-holder and not a digitization hunt, and the State Library of Ohio is the obvious first one: it is the institution most likely both to hold the subscription and to answer a reference request. The corpus must not record the era as unreachable in general — what is established is that one publisher does not serve it and another does, for a fee.

Two routes remain genuinely unprobed: the Ohio History Connection, and the department’s own historical annual reports. Both are questions to an institution rather than URLs. See #18.