Data Sources

What Each Standard Contributes

These are the standards whose wall tables are published here, one row per document rather than one per page. The counts are the data itself: how many rows exist under each standard, and how far each row has been checked.

One document has no verified rows at all — GOST 8732-2025. Nothing in it has been corroborated a second time, so every row it contributes is labelled single source and none of it may be presented as confirmed. That is also why it shares a page with the two GOST standards that do have verified rows, rather than getting a page of its own.

The ASME B36.10M-2015 table published at ASME B36.10M steel pipe dimensions and weights is the same standard again, keyed by nominal size and schedule instead of by diameter, because it is the one document here that publishes a nominal size column.

How Each One Was Obtained

Extraction method matters more than it sounds. Every table here was parsed deterministically — regular expressions over raw text, or x-coordinates over a PDF — and no language model was allowed to read a table for numbers.

The ASME row is the one that changed everything. The standard’s own Table 2 extracts cleanly with pdftotext -layout, which turned the verification basis from “eight publishers agree” into “the standard says so” — and immediately exposed four nominal sizes our parser had been dropping silently, about which all eight publishers had agreed with us because none of them had ever been asked.

Raw source documents are retained rather than discarded, so every number on this site can be re-derived from the file it came from instead of taken on trust. How this data was verified, and the gaps that remain describes the checks each figure then has to pass.

What Is Paywalled, and What Was Used Instead

Some of these documents sit behind a paywall; others are nominally available and still could not be read. Naming them is the point: a reference that quietly substitutes a distributor table for a standard is making a claim it has not earned.

GB/T 17395-2024, Tables 1 to 4

Pages 8 to 47 are Private-Use-Area glyphs with no character map, and the standards body serves encrypted image tiles. Roughly US$800 to close. Published here instead: the 2008 edition, named as such on every table, because that is the edition that could be read.

KS D3507 and KS D3562, the Korean standards

Paywalled. The Korean figures in the wider research rest on distributor and engineering-reference tables rather than the standard, so no Korean wall table is published here at all.

AS 1074-1989, Australia

Behind the Standards Australia paywall. Four publishers that agree cell for cell, plus the standard's own published abstract, stand in its place.

IS 8329, Indian ductile iron

Access denied twice. Sizes DN 600 to 1200 rest on a single publisher and are labelled accordingly.

EN 1057 Table 3, the copper wall matrix

The only freely available copy rate-limited, and its cell grid collapsed on extraction. The outside diameters and tolerances survived; the matrix did not, and it is not published.

US HDPE DIPS walls, DR 17 to 32.5

Not paywalled — simply wrong. The only source found fails its own wall-from-diameter relationship, so it was rejected outright rather than published with a caveat.

AS/NZS 1477 Series 2 walls

One publisher. Held at single source until a second appears.

Where the Evidence Stops

Three gaps cannot be closed by looking harder, and all three are published rather than papered over.

Eight single-digit typos in GOST 8734 and GOST 10704. No upstream correction exists. GOST 8732 was reissued in 2025 with all eleven of its deviations fixed; 8734 has had no such reissue, and every accessible online copy descends from the same scan — so a second look returns the same typo and reads like confirmation.

The unnumbered ASME wall thicknesses, the majority of that table. One publisher lists them, corroborated by that same publisher’s separately-transcribed mass column through the formula check. That is a genuine consistency check and it is not a second publisher: sound to compute with, not safe to quote as authoritative one row at a time.

Source count is not source independence. Five clusters of apparently separate sources turned out to be the same content republished — nine domains serving a byte-identical template, three sites reproducing one optical-recognition master down to a defect in a column header, and three widely-cited references faithfully reproducing the same erratum from the standard they had copied. Three sources agreeing, all wrong. Errors found in published pipe data, and in ours lists what that has caught.

The Arbiter Is Arithmetic

Where sources conflict, the tiebreak is not a vote. Every standard publishes a relationship between outside diameter, wall thickness and mass per unit length, and recomputing the published mass tests all three numbers against each other at once:

M = (OD - t) * t * C

The constant is material-specific and unit-specific. C is 0.02466 kg/m with millimetres, or 10.69 lb/ft with inches, for carbon steel at 7.85 g/cm³. Ductile iron is 0.0221482, and using the steel value on iron overstates the mass by 11.3%. There is one further trap the wall dataset records against itself: the ISO 65, EN 10255, AS 1074 and IS 1239 families compute mass from the mean of the outside diameter tolerance band, not from the nominal outside diameter, so applying the plain formula to those tables produces a small, systematic and entirely avoidable error.

Revision

Wall tables compiled 2026-08-16, data revision 2026-08-16. Corrections are dated and published rather than applied quietly, and the standards themselves are compared side by side at the pipe standards published here and where they disagree.