The short version.

• India's central gazette publishes tens of thousands of notifications (G.S.R./S.O. rules) with no official cross-reference index — nothing on egazette.gov.in tells you that a notification you're reading amends an earlier one, or what later superseded it. An independent research project (Gazette Trail, source on GitHub) set out to build that index by hand, ministry by ministry — and has since run a much deeper pass than this piece first covered: 2,591 notifications and 2,113 cross-references across all 53 central ministries and departments, up from 400/285 at first publication. This piece has been amended to reflect that expansion, independently re-verified rather than taken on the project's word (§10).
• Try it yourself: the Gazette Trail Reader, this piece's companion search tool, now covers all 2,591 notifications (§2).
• The biggest number in the whole dataset: the Indian Post Office Rules, 1933 were amended 207 times between 1975 and 2022 — a 47-year history recited in full inside a single 2023 notification's own "Note," now modeled as 206 distinct linked notifications (§4). It dwarfs this piece's original lead story.
• That original lead remains a genuinely strong, independently-corroborated story on its own terms: the Ravi and Beas Waters Tribunal's report deadline has been extended by gazette notification 36 times since 1986 — still unresolved, still running (§3).
• One office quietly governs everyone else's paperwork: the single most-cited notification in the entire corpus is a 1986 rule on which government bodies fall under the Central Administrative Tribunal, cited 26 times since; the Department of Personnel & Training holds 455 cross-references overall, more than double the next-busiest ministry (§6).
• This piece independently rebuilt the project's database again from its own published source files, the same discipline applied at first publication. This time the numbers match almost exactly — 53 ministries, 2,591 notifications, 2,113 cross-references, down to the exact relation-type and provenance-tier breakdowns — with one small, fully explained file-load-order quirk, not the two-file gap found at first publication (§10).

The invisible link: what a single gazette notification can't tell you

A Government of India gazette notification — a G.S.R. (General Statutory Rule) or S.O. (Statutory Order), the everyday instrument by which a ministry amends a rule, extends a deadline, or reconstitutes a committee — is published as a standalone document. Read one on egazette.gov.in and you get exactly what's on the page: a date, a rule text, sometimes a "Note" reciting its own history. What you don't get, anywhere in the official record, is a queryable answer to "what did this amend, and what amended it before that." Gazette Trail, the project this piece is built on, set out to construct that missing index from scratch: a small SQLite database (gazette_notification and cross_reference tables, with an amends | supersedes | corrigendum | rescinds | repeals | cites relation type on every edge) populated one real, dated, individually-researched notification pair at a time, across every central ministry and department.

The method, as documented in the project's own files, is bounded and disclosed rather than exhaustive: most notification text was researched via gazettetracker.com's Full Text search (a third-party aggregator, not the official egazette.gov.in archive itself), each finding transcribed by an AI research pass; a meaningful minority (18% of all cross-references, see §10) were instead independently opened and read directly from egazette.gov.in's own PDFs, a higher-confidence tier the project itself distinguishes rather than blends in. A citation is only modeled when it cites a G.S.R./S.O./S.R.O. number on both ends; notifications citing a bare Act, Regulation, or file number — a real and common citation form in Indian drafting — are deliberately left unlinked rather than forced into a shape the schema doesn't fit. Every one of those choices matters for how much weight the numbers below can bear, and this piece flags them again in §11 rather than letting the topline findings imply more certainty than the underlying research supports.

A tool to actually read one: the Gazette Trail Reader

A citation chain described in prose, however precisely, is still something a reader has to hold in their head. This piece also publishes a small standalone tool built from the same rebuilt database: the Gazette Trail Reader. Search any of the 2,591 notifications in this dataset — by ministry, by G.S.R./S.O. number, by subject, or by year — and see, in plain language, what it amends, supersedes, corrects, rescinds, or repeals, and what later happened to it. It is meant as an accessibility aid in the plain sense of the word: the underlying information already exists in the gazette, scattered across individually unremarkable-looking documents; this tool just makes the connections between them readable without requiring a reader to already know where to look. Try searching "Ravi Beas" or "Post Office" to see the two chains in §3 and §4 laid out interactively, or browse by ministry to see what this dataset covers before reading further. The tool works entirely in your browser (nothing is sent anywhere), and its underlying data is separately downloadable as plain JSON for anyone who wants to build something else with it.

The tool carries the same sourcing caveats as this piece throughout: coverage is a bounded research sample (53 ministries, on this piece's own from-scratch rebuild — see §10), and most links are transcribed by an AI research pass from a third-party aggregator rather than independently re-checked against the primary PDF (a growing minority now are — see §10's provenance-tier breakdown). The tool's own footer repeats this on every visit rather than assuming a reader arrived via this article.

36 extensions, 40 years, one tribunal that has never finished

The Ravi and Beas Waters Tribunal, constituted by a notification dated 2 April 1986 under the Inter-State River Water Disputes Act, 1956, was this piece's original lead story, and remains one of the best-corroborated chains in the whole dataset even though (per §4) it is no longer the longest. It adjudicates how much Ravi and Beas river water Punjab, Haryana and Rajasthan each get. The tribunal submitted an initial report in 1987, but a further reference back to it — on the volume of water Punjab has "surplus" to share — has never been resolved. Every time the tribunal's deadline to report back approaches, the government gazettes a fresh S.O. extending it. The project's database traces this as one continuous 36-node chain: notification after notification, from 1986 through a notification dated 11 June 2026, each one naming the extension before it.

This piece independently searched for the tribunal rather than taking the project's chain on faith, and found real, independent corroboration — strong enough, on a closer pass, to resolve a discrepancy this piece originally flagged rather than leave it open. Press coverage confirms the tribunal, its 1986 constitution, its 1987 initial report, and describes a deadline extension "attributed to the exigencies of the work involved," issued by the Ministry of Jal Shakti in July 2025, pushing the deadline "by one year" to 5 August 2026. That matches the project database's 35th entry — S.O. 3056(E), dated 9 July 2025 — almost exactly, both the date and the extension language. This piece's first pass at this comparison mistakenly checked the press coverage against the database's 36th and latest entry (S.O. 3024(E), 11 June 2026) instead, which produced an apparent date mismatch that was never a real discrepancy — it was comparing the wrong two links in the same chain. Once compared correctly, the 36th entry reads exactly as the pattern predicts: a further extension, issued in June 2026, ahead of the 35th extension's own 5 August 2026 deadline, continuing the near-annual pattern that has run since 1986.

The real record-holder: 207 amendments to one 1933 rulebook

The project's later depth pass surfaced something that makes even the Ravi-Beas tribunal look modest: the Indian Post Office Rules, 1933 — the foundational rulebook governing how India's postal service actually operates — were amended 207 separate times between 19 December 1975 and 29 December 2022, a 47-year span. Nobody had to piece this together notification by notification: a single 2023 notification, G.S.R. 17(E) of 10 January 2023, recites the entire history as one enumerated "Note," which the project transcribed in full and modeled as 206 distinct linked notification rows (the 207th listed item is the Finance Act, 2017 itself — an Act of Parliament, not a gazette notification, so no notification-to-notification edge bridges it, an honest gap rather than an invented one).

Two things make this chain a genuine stress test of the whole project's method, handled transparently rather than smoothed over. First, the same G.S.R. or S.O. number in this list legitimately identifies different, unrelated notifications years apart — 462(E) in 1988 and again in 2023, 639(E) likewise, 379(E) three separate times — because numbering resets periodically; every collision was checked individually and kept as distinct rows rather than merged. Second, at least one entry in the Note gives only a month and year ("December, 1989"), forcing a placeholder day rather than a fabricated precise date. Both are disclosed in the project's own source comments rather than silently resolved. Separately, the newer Post Office Regulations, 2024 — a different, more recent rulebook under the same department — already has its own 15-notification amendment chain running December 2024 to August 2026, seven amendments in 2025 and seven more in 2026, suggesting this pace of continuous, granular amendment is simply how the Department of Posts operates rather than a one-off.

The 207-entry Note and its two disclosed data-quality issues (number-reuse across years, one month-only date) are as documented in the project's own db/seed_posts.sql file; this piece did not independently open the citing 2023 notification's own PDF to re-verify the transcription, and reports it at the project's own research-agent-quoted/primary-source-full-text confidence mix rather than a higher tier.

A notification published twice, then quietly withdrawn

A smaller but sharper example, from the Department of Telecommunications side of the Ministry of Communications: on 21 November 2024, the ministry gazetted the Telecom Cyber Security Rules, 2024 (G.S.R. 720(E)). Amendment Rules followed on 22 October 2025 (G.S.R. 771(E)). Then, per the project's own research notes, the same amendment appears to have been gazetted a second time under a different number, G.S.R. 796(E), published a week later on 29 October 2025. A third notification, G.S.R. 863(E) dated 25 November 2025, then rescinded the duplicate — formally withdrawing 796(E) and leaving 771(E) as the version actually in force. None of this is visible from reading any one of these four notifications in isolation; a reader who found G.S.R. 796(E) on its own, without knowing it had been rescinded a month later, would have no way to know it no longer applies.

One office quietly governs everyone else's paperwork

Rank every notification by how often other notifications cite it, and one office dominates in a way nothing else in the dataset comes close to. The single most-cited notification in the entire corpus is G.S.R. 730(E), issued 2 May 1986 by the Department of Personnel & Training (DoPT) under section 14(3) of the Administrative Tribunals Act, 1985 — it lists which government bodies and corporations fall under the Central Administrative Tribunal's jurisdiction. It has been amended 26 times since, almost every amendment simply adding another body to the list as new corporations and authorities are created or reorganized — and unusually for this dataset, most of those 26 edges are tagged at the project's highest confidence tier, primary-source-egazette (independently opened and read, not just quoted from an aggregator). A second DoPT rule, G.S.R. 152 of 8 September 1954 (the IAS/IPS Cadre Rules), is cited 23 times more, as individual states' cadre-strength notifications reference back to it decade after decade.

Zoom out to whole-ministry totals and the same pattern holds: DoPT (folded into the Ministry of Personnel, Public Grievances and Pensions) is the source of 455 cross-references — more than double the next-busiest ministry, Communications (246), and well ahead of Finance (147), Ports & Shipping (113), and Jal Shakti (84). If one single thread explains how the Indian civil service actually gets regulated day to day, on this dataset's evidence it runs through DoPT's decades-old central service rules being patched and re-patched — not through any individual ministry's own founding legislation.

How ministries differ: patch it, or replace it whole

Across the full 2,113-edge dataset, amends outnumbers every other relation more than ten to one (1,621 of 2,113 edges, 76.7%) — Indian subordinate legislation is overwhelmingly incremental rule-maintenance, a clause changed or a schedule updated, not wholesale replacement. Supersedes (155) and rescinds (32) together account for under 9% of all edges, reserved for genuine resets. Corrigenda — formal typo-fixes to an already-published notification — number 90, common enough to count as a real drafting sub-genre rather than noise.

But that national average hides real ministry-level variation. The Ministry of Ports, Shipping and Waterways is the only ministry in the dataset where full replacement genuinely outpaces incremental fixing: 45 supersedes edges against just 20 amends, driven by Major Port Authority boards that reissue their entire membership list rather than amend a name in it. The Ministry of Finance sits at the opposite extreme: 101 amends against just 33 supersedes, reflecting Customs' habit of endlessly patching one consolidating notification rather than ever replacing it outright. Mines (40 amends vs. 8 supersedes) and Social Justice and Empowerment (38 amends vs. 6 supersedes) show the same patch-don't-replace instinct at scale, chaining Concession Rules and RPwD/NCSK reconstitutions repeatedly rather than starting over.

Two real institutional resets from this same period landed in the dataset as edges rather than being silently overwritten by a "current rules" snapshot, which is the point of tracking supersession explicitly: MGNREGA's wage-rate notifications were repealed outright on 1 July 2026, folded into a new Act's own wage-rate provision, with both threads marked superseded rather than deleted from the record; and the Science and Engineering Research Board (SERB) was rescinded and absorbed into the new Anusandhan National Research Foundation (ANRF), its 2014 staff regulations superseded by ANRF's own recruitment rules. A static snapshot of "current rules" would show neither of these as anything — the old rule would just vanish. Tracked as a graph, the transition itself is the queryable fact.

An honest gap, kept honest

Not every long chain in this dataset is fully documented end to end, and the project's own discipline around that is worth a specific example: Rural Development's MGNREGA Schedule I wage-rate amendments form a 14-node chain from 2013 to 2025, each notification's "Note" naming its immediate predecessor — reaching back to a 2022 notification. Separately, that same Note names a 2007 notification (S.O. 323(E)) as Schedule I's very first amendment. But no single quoted text bridges the 15 years between 2007 and 2022, even though real intervening amendments plainly exist — MGNREGA's wage Schedule is amended often. Rather than guess at the missing links, the project recorded 2007 as a real, dated fact and left no edge connecting it to 2022. A documented edge in this dataset means a specific quoted citation was found; an undocumented gap means real links likely exist but weren't individually captured, not that nothing happened in between — the same discipline visible in the Post Office chain's own honestly-unbridged 207th entry (§4).

When a committee's gazette trail crosses a ministry reorganization

One more finding worth isolating: what happens to a citation chain when the government itself reorganizes. The National Cooperative Development Corporation's General Council membership was last set by a notification issued in July 2020 — under the Ministry of Agriculture and Farmers Welfare, because the Ministry of Cooperation did not yet exist as a separate ministry. It was created in 2021. When the General Council's membership was next revised, in February 2024, the amending notification came from the new Ministry of Cooperation — a different ministry, on paper, superseding a notification from a ministry that, organizationally, no longer has anything to do with this subject. The project modeled this correctly (the 2020 notification stays recorded under Agriculture, its real ministry_id at the time; the 2024 one is Cooperation's own), rather than either inventing a Cooperation-era ministry for 2020 or silently dropping the cross-ministry link. It's a small, specific illustration of a structural feature of India's gazette record: a machinery-of-government change orphans the citation trail for anyone not already tracking it across the boundary.

What this piece verified itself, rather than took on faith

The project's own final status file claims, as its closing gate for this depth pass: "53 ministries, 2591 notifications, 2113 cross-references; 141 tests passed" with a foreign-key integrity check passing cleanly. As at first publication, this piece did not take that claim at face value — it re-cloned the repository and rebuilt the database itself, in Python, from the project's own schema.sql plus every db/seed_*.sql file.

This time the rebuild matches almost exactly, a meaningfully cleaner result than the two-file gap found at first publication. Loading all 53 seed files in plain alphabetical order, as the project's own documented process specifies, only one file fails: db/seed_ayush.sql, because one of its own notification rows models a real cross-ministry supersession — Ayush's S.O. 3559(E) superseding an erstwhile Ministry of Health and Family Welfare notification, correctly modeled under Health's own ministry_id — and "Health" sorts alphabetically after "Ayush," so Health's ministry row doesn't exist yet when Ayush's file loads first. Loading seed_ayush.sql last instead of in strict alphabetical order (the same kind of ordering fix, on a smaller scale, that the two-file gap at first publication needed) resolves it completely: 53 ministries, 2,591 notifications, 2,113 cross-references, zero foreign-key violations — an exact match to the project's own claimed totals.

The project also tags every cross-reference with its own provenance tier, and this piece verified that breakdown directly rather than repeating the project's summary uncritically: 1,726 edges (81.7%) are research-agent-quoted (an AI research pass's own transcription of a real quote from primary text, sourced via the aggregator); 380 edges (18.0%) sit across several higher-confidence primary-source tiers — independently fetched and read directly, whether the full notification text, the raw egazette PDF, or just its preamble; and a final 7 edges (0.3%) are explicitly lower-confidence (title/date-resolved rather than a full quote, or flagged uncorroborated). That tier distribution is exactly where an outside check should start, and this piece's own re-derivation of it, cross-reference by cross-reference from the raw database rather than copied from the project's summary page, matched the claimed figures exactly.

One more real finding, also independently checked rather than assumed: the project's coverage plan documents 3 ministries — Law and Justice, Tourism, and Panchayati Raj — as deliberately abandoned after a genuine research pass turned up nothing linkable. Law and Justice's real amendment pairs all cite Act or Regulation numbers, never a G.S.R./S.O./S.R.O.; Tourism's tracker coverage runs to only 4 notifications total, each amendment citing a G.S.R. number on one side only; Panchayati Raj has just 2 tracked notifications, both standalone Resolutions with no relationship between them. That's a genuine finding about how those ministries' delegated legislation is drafted and numbered, not a gap in the research.

What doesn't follow from any of this

Every number and chain in this piece describes what one bounded, if now much deeper, research pass, using one third-party aggregator plus a growing share of direct primary-source reading, found and could transcribe into the citation shapes this project's schema models — not a complete or official record of India's gazette history. A ministry modeled in the project's coverage plan means "captured what this pass turned up," not "modeled everything that exists"; a longer or shorter real chain may well exist for any subject covered here that this research simply didn't surface, and the 207-entry Post Office chain in §4 is itself evidence that a single later pass can still surface something that dwarfs everything found before it. The 36-node Ravi-Beas chain is independently corroborated against outside press coverage on its most specific, checkable claim (the 35th extension's exact date and language — §3), but most other chains in this dataset rest on the project's own transcription alone, at whichever provenance tier it's tagged; this piece did not independently re-verify each one against a primary PDF itself. The database-rebuild quirk in §10 is a single, small, fully explained file-load-order issue — a real improvement over the two-file gap found at first publication — and says nothing negative about the correctness of the roughly 2,591 real notification records that do load and check out cleanly.

Sources and caveats