search

w3c / vc-data-model

W3C Verifiable Credentials Working Group — VC Data Model and Representations specification
https://w3c.github.io/vc-data-model/
Other
281 stars 98 forks source link

Updated section on semantic interoperability (issues #1217 and #1210) #1420

Closed iherman closed 4 months ago

iherman commented 5 months ago

This PR is an attempt to jointly cover issues #1217 and #1210, by a thorough re-write of §5.3.1 on Semantic Interoperability.

This is a normative section, ie, special attention should be given to the usage of the MUST, SHOULD, etc, keywords. My general approach was to reduce the MUST to the strict minimum, but reviewers might differ.

Preview | Diff

iherman commented 5 months ago

I don't know about the testability of all the MUSTs here and whether, on account of that, they should instead be "expectations" or recommendations.

The change could be to use SHOULD/RECOMMENDED everywhere.

However. "Testability" in the W3C CR setting is a very general concept; the goal is to prove that a feature is implementable by two, independent implementations (to ensure interoperability). From a CR point of view, this does not mean that each MUST should have a running test program. (The usual counterexample is the "testing" of a vocabulary, which means that the terms are in use by at least two different implementations.) Taking this view, the few MUST statements in this section are easily testable: it requires at least two VC applications that define their own application specific terms, and that abide to these requirements. Isn't that feasible?

(Personally, I would prefer, from the spec's point of view, to keep those MUST-s.)

iherman commented 5 months ago

@TallTed:

Editorial...

All adopted.

iherman commented 4 months ago

Some of the normative statements feel a bit heavy-handed (and untestable -- or should we test them?).

See my comments in https://github.com/w3c/vc-data-model/pull/1420#issuecomment-1918409250.

If we do have two "properly defined" vocabularies around as part of the implementations, then, CR-wise, we are o.k. No reason for explicit "tests" imho.

iherman commented 4 months ago

Overall, we should decide if we're also addressing issue #1410 with this PR (as we're dangerously close to doing so, IMHO).

We are not "dangerously close": we are covering it. A 'type' is part of a vocabulary, whether new properties are defined or not.

I am not sure whether we need to make it explicit in this text, and how. Maybe by defining (in the terms' section) what we mean by 'vocabulary': something that may be as simple as the definition of new types, and as complex as defining whole new properties.

iherman commented 4 months ago

The issue was discussed in a meeting on 2024-02-07

View the transcript #### 3.4. Updated section on semantic interoperability (issues #1217 and #1210) (pr vc-data-model#1420) _See github pull request [vc-data-model#1420](https://github.com/w3c/vc-data-model/pull/1420)._ **Manu Sporny:** A number of PRs for VCDM. Huge thanks to ivan for raising a number of them. One we should talk about today (1420). Ivan raised. … a question remains. We just entered CR. I think this PR is normative. It does not impact software implementations but does impact vocabulary. Wonder if this triggers a second CR. **Ivan Herman:** AFAIK, a new draft with normative changes are OK provided that at some point in time we publish a snapshot with these changes. I don't think there is any problem with these few normative changes here. We know we will republish a snapshot sometime in May. I think we are fine. > *Brent Zundel:* +1 to Ivan. **Manu Sporny:** That is my read as well. I've started tracking this with a normative label so we know we have done a second CR triggering thing and that we can summarize these changes. If you raise a PR, please try to classify as Editorial or Normative. … also keep in mind that any normative change needs to be communicated with the test suite workers. **Ivan Herman:** If a change is normative in a serious way, then the editor of the PR should add an item to the list of changes at the end of the document or we will forget them. > *Manu Sporny:* yes, +1 to that. **Brent Zundel:** this is something reviews should remind when examining PRs. … Any other core data model status? **Manu Sporny:** A number of editorial changes are out there and folks should pay attention.
msporny commented 4 months ago

Normative, multiple reviews, changes requested and made, no objections, merging.

msporny commented 4 months ago

BTW, I think this PR just created a new conformance class for vocabulary documents in the spec... If you agree, @iherman, we might want to add that conformance class and point it to the semantic interop section. I'll also note that I still think that this is a partial solution for #1410 because we don't add the text that I believe @jyasskin was looking for.