Semantic edition diffs
The service states what changed between two editions at the level of requirements and limits, rather than as text differences.
Each job below states the situation, what the service does, and what the member sees, together with an example question and the assets that power it. Where a job can be demonstrated deterministically, the demonstration runs on this page.
Metanorma+ Glossarist+ Relaton+ Pubid
"Where does R 60 address creep, and what is the maximum load for creep testing?"
A member asks a question in their own words, at any hour, about any clause in the library — and needs an answer they can act on.
What the service does — The service reads the question, resolves everyday phrasing to the corpus's defined terms, retrieves over the clause-structured corpus, and generates the answer under a contract that requires a citation on every claim and verbatim quotes for normative values.
What the member sees — An answer whose every claim names the publication, edition and clause behind it; a citation opens the original document at the cited clause, with the clause highlighted.
The rendered form
An answer arrives with inline citations such asR 60-1:2021 §4.4.2 on every claim; normative values appear as verbatim quotes; a citation opens the original document at the cited clause, with the clause highlighted.
See the full anatomy → Home+ Primmel
"Is a test cell with D_max = 26 000 v valid for E_max = 30 000 v?"
A member states an instrument's values and asks whether the configuration conforms to the standard — a question with a computable answer.
What the service does — The service binds the stated values to the rule's parameters, evaluates the check deterministically, and attaches the verdict to the answer as data, with the arithmetic shown and the standard's own word for any violation.
What the member sees — A verdict block — valid, invalid with the reason in the standard's own words, or void naming what is missing — computed rather than judged, which the prose must present faithfully.
D_max ≤ 0.8 · E_max0.8 × Emax = 24 000 v
Dmax / Emax = 0.867
exceeds the limit by 2 000 v
INVALID
This constraint comes from a publication model in the reference deployment’s corpus, and it is computed here in your browser — the same deterministic evaluation the service attaches to answers as data, with the standard’s own word for the violation. Hypothetical values are evaluated the same way as real ones.
+ Primmel
"Does R 60 constrain packaging?"
A member asks whether the standard constrains something, and "we found nothing" is not good enough — they need to know that nothing is there.
What the service does — The service enumerates the standard's entire machine-readable model — every requirement, constraint, test and term — and returns a certificate of absence.
What the member sees — A certificate stating how many nodes were examined, that none matched, and what scope the enumeration covered.
Certificate of absence · illustrative
query does the standard constrain packaging?
enumerated requirements constraints tests terms
result 0 matches in N nodes
scope the machine-readable model, in full
conclusion the property is absent, by enumerationMetanorma
"Check this draft answer against the corpus."
A member has an answer — from this service, from another system, from a colleague — and wants to know whether it holds against the corpus.
What the service does — The deterministic checks verify that each quoted span appears verbatim in a served passage and that every referenced object resolves; a separate judged check scores faithfulness, labeled as judged.
What the member sees — Mechanical and judged results, each labeled as what it is, for any answer submitted to the check.
Passage · served excerpt (illustrative)
The load-cell family shall be designated by its maximum capacity E_max. The test load D_max used for creep testing shall not exceed 0.8 · E_max.
Claims under the contract
quote found in passage — answer served
Both checks here are real substring containment, the same shape as the deterministic battery: the quoted span must appear verbatim in a served passage. Judged components in the service are labeled as judged — this one is mechanical.
The service states what changed between two editions at the level of requirements and limits, rather than as text differences.
When a limit changes, which requirements, tests, verdicts and forms change with it — change impact over the model's dependency graph.
Ordered test sequences and evaluation processes as queryable structure: what gates what, what runs next, what is contaminated by what.
The frontier is measured before it ships; its definition and current state are onthe annealment page.