Does MTPROTO replace our existing certificate templates?
No. Most laboratories that have adopted it emit a record alongside the document they already produce, with the same content. The PDF stays the thing a person reads and an auditor files; the record is the thing a system reads. Nothing about the accreditation picture changes.
How does this relate to the existing digital calibration certificate work?
It overlaps, deliberately and narrowly. The larger XML-based efforts model a complete certificate, including everything a national metrology institute needs to express. MTPROTO models the subset that a receiving system actually acts on: what was measured, how well, and against what. It is straightforward to derive an MTPROTO record from a richer certificate, and by design not possible to go the other way without loss.
If the full model fits your workflow, use it. MTPROTO exists for the many organisations for which it does not, and which are currently exchanging scans.
Why another format instead of extending one that exists?
The working group spent most of 2019 trying to do exactly that. The blocker was not technical goodwill but versioning: a change to a certificate schema that a dozen national institutes depend on takes years, correctly so. A small format with a sixty-day review period can move at the speed of the instrument vendors who need it.
Is a signature required?
No, but an unsigned record is only as trustworthy as the channel it arrived on. In practice signing is cheap — the key material is 32 bytes and verification needs no network access — so most producers sign everything. See §7.
What happens when a key is compromised?
Nothing happens to the signatures on existing records, because verification checks the key named in the record. What changes is whether you still believe the records that key signed. That is an organisational judgement, and the mechanism for expressing it is the issuer publishing superseding records — not an online revocation list this specification would have to define and everyone would then have to reach.
Can I add fields we need internally?
Yes. Unknown members must be preserved by conforming consumers, so a record carrying
your internal job number survives a round trip through someone else's system intact.
Use an x- prefix for private quantity names to avoid colliding with a
future registered one.
Is there a conformance test suite?
There is a corpus of about four hundred records — valid, invalid, and
deliberately awkward — used by the reference validator. It is published with
mtp-validate. It is not a certification programme, and the working group
does not issue conformance marks.
Who decides what goes into the specification?
Whoever turns up. Proposals are made on the list, normative changes get at least sixty days of review, and the editors' job is to write down what the discussion concluded rather than to arbitrate it. The bar for adding anything is a described problem and at least two parties who have it.
Under what licence is all this published?
Specification text and schema files are CC BY 4.0. Reference tooling is MIT. There are no patents asserted over the format by anyone in the working group, and participation is conditional on not asserting any.