Defining implementation conformance scope

onthebreeze commented 5 months ago

This ticket is to provoke discussion about what it really means to "implement" UNTP

There will be different versions. We should follow semver (semantic versioning) major.minor.patch numbering with only major version changes representing breaking changes. So any v1.x implementations should be interoperable.
There are a group of independent component specs. Do we version then independently? Probably yes. But do we assess & claim conformity at the granularity of a dozen lower level specs? Could be confusing. Perhaps some grouping mechanism.
On top of that, every component spec will have IETF MUST, SHOULD, MAY language to indicate what is mandatory and what is just recommended or nice to have.
And finally not al components of each spec will be relevant for every actor role. For example conformity assessment bodies will just issue conformity credentials and perhaps need have nothing to do with digital product pasports or traceability events etc.

Here's a little diagram with some thoughts on this. Lets discuss.

UNTP_Implementations

NancyNor commented 5 months ago

hi all, if we have different levels of implementation (minimum, recommended, full) who is the body that evaluates the level of each implementer? Are we resourced to do these evaluations on an ongoing basis? Is this evaluation automated somehow? Maybe it should just be a pass/fail to keep things simple.

nissimsan commented 3 months ago

Section: "This is How to Claim Conformance".

Requires test suites.

nissimsan commented 3 months ago

Schema compliance. Data model-focused.

nissimsan commented 3 months ago

@onthebreeze , please share your mental model in a PR!

seewodg commented 3 months ago

Perhaps if the concept of conformity has various scopes it could be valuable to consider "profiles" as in: https://www.w3.org/TR/dx-prof/

philarcher commented 1 month ago

Section: "This is How to Claim Conformance".

Requires test suites.

As a general principle: if there's a MUST or a SHOULD in the spec then, ideally, it's something that can be tested in a test suite. If it can't be tested either by a machine or by a human using some objective criteria, maybe it shouldn't be a MUST??

uncefact / spec-untp

Defining implementation conformance scope #14