Updated 28 September 2026

1.6.0 - 2026-09-28

• Added 23 records: 14 low-cement and dense castables, 9 insulating castables and boards. The castable work closes a gap that had been open since v1.2.0, when it became obvious that monolithics were being treated as an afterthought.
• k_method is now required whenever k_1000_w_mk is present. 46 existing values turned out to be calculated rather than measured and are flagged as such; 3 had no determinable method and the conductivity was withdrawn.
• k_ref_c populated on every record that carries a conductivity, replacing the previous convention where 400 °C figures were silently mixed with 1 000 °C ones.
• Four sources added, numbered S-036 to S-039, three of them conference papers on spinel-bearing castables. That brings the set to 39 issued ids and 38 live ones.
• Corrections: CC-0103 alumina content 84.1 → 85.2; CC-0241 conductivity withdrawn as unsourced; CC-0292 density 1.52 → 1.55.
• Credited: A. Rezaei, M. Dlamini, and two submitters who asked to stay anonymous.

1.5.1 - 2026-08-11

• Eight insulating firebrick densities were a factor of ten out, the result of a lb/ft³ conversion applied twice during the v1.2.0 import. Reported by a reader who noticed that a class 23 brick appeared denser than water.
• The conversion table in the methodology page now states the rounding rule for each factor, because the double application had been invisible to review.
• No schema change, no records added or removed.
• Credited: J. Lindqvist.

1.5.0 - 2026-06-30

• Not backwards compatible. thermal_shock_cycles kept its name but lost its old meaning: the protocol moved into the new shock_protocol field, and every existing count was re-examined against its source. Nineteen counts could not be attributed to a protocol and were emptied.
• Three protocol values defined: water quench at 1 000 °C, forced air quench at 1 000 °C, prism test at 950 °C. Mixing counts across them had been the dataset's worst silent trap.
• Added 23 records across the magnesia and magnesia-carbon families.
• cinderq 0.9.0 released alongside, with diff added specifically so this change could be audited.

1.4.0 - 2026-03-17

• Chrome-bearing family taken from 6 records to 22, including rebonded magnesia-chrome and the chrome-corundum grades used in waste incineration linings.
• cr2o3_pct added as a field. It is empty outside this family rather than zero, since most sources simply do not report it.
• Two handbooks added as sources, S-034 and S-035. One 1978 compilation was withdrawn after its figures turned out to be reprinted from an older document already in the set; its id S-021 is retired and will not be reused.
• Credited: A. Rezaei.

1.3.0 - 2025-12-02

• hmor_mpa introduced, populated for 71 records. The field ships without a test temperature, which was flagged as a known weakness on the day it landed and is still open.
• Added 31 records, mostly high-alumina bricks between 50 and 80% alumina.
• Family boundary clarified: alumina-spinel grades sit in corundum, not magnesia, because that is where readers were looking for them.

1.2.0 - 2025-08-19

• Insulating family extended to fibre blankets, boards and microporous panels: 28 records, all with conductivity at 400 °C because no source quoted 1 000.
• That decision created the mixed-reference-temperature problem that k_ref_c eventually fixed in v1.6.0, and the lb/ft³ bug that v1.5.1 fixed. Both were introduced here.
• Added 28 records in total, plus two association compilations as sources, S-032 and S-033.

1.1.0 - 2025-04-28

• Every source note rewritten to state the limitations of the document rather than just its bibliography.
• Fourteen journal and conference papers added, taking the source count to 31 and the record count to 282.
• CSV output switched to LF line endings and UTF-8 without a byte order mark, after a reader reported that a spreadsheet was showing a stray character in the first header cell.

1.0.0 - 2025-01-20

• Schema frozen. Record ids became permanent and the renumbering that happened twice during 0.x will not happen again.
• 254 records, 22 fields, 17 sources.
• Licensing settled on CC BY 4.0 for the data and MIT for the reader, replacing the earlier and vaguer "free to use" wording.

0.9.0 - 2024-10-07

• Confidence flag added, and the rule that created it: no method, no number. 63 values collected from manufacturer literature without test conditions were removed in one pass, taking the record count down before it went up.
• Reached 198 records. The purge is the reason 0.9.0 took eight months after 0.5.0.

0.5.0 - 2024-02-13

• Not backwards compatible. All imperial units converted to SI and the original figures moved into source notes. Field names changed to carry their unit, which is why k_1000_w_mk looks the way it does.
• 121 records, 9 sources.

0.2.0 - 2023-09-05

• Family enumeration settled at eleven values after three rounds of argument about whether monolithics deserved their own family. They do not; they sit with the chemistry they are made of.
• 64 records, first JSONL output alongside the CSV.

0.1.0 - 2023-06-14

• 31 records transcribed from two handbooks, published as a single CSV with a README and no schema. Started in March 2023 as a shared spreadsheet between two engineers tired of re-deriving the same table.