Address normative concept of VerifiableCredentialGraph

[OR13]

See the original comment here: #1236 (comment)

The word "envelop" is being used here, in a way that aligns with "VerifiablePresentation".

I find the concept of "VerifiableCredentialGraph" confusing, and note that it does not occur even once in the current TR.

https://github.com/search?q=repo%3Aw3c%2Fvc-data-model%20VerifiableCredentialGraph&type=code

I prefer to remove the term, since it can only lead to more confusion.

If the term is providing value, I prefer to unify that value with "VerifiablePresentation", but would need to know more about how it actually works, to even propose a path to that.

[OR13]

In the original comment, @iherman wrote:

How is it different from: the former is the "envelope" that contains the latter. Just as the envelope is not the same as what it contains, the VerifiableCredentialGraph is different from what it contains, i.e., a VerifiableCredential. This differentiation is crucial for, e.g., the usage of the RDFC 1.0 algorithm.

Can someone elaborate further on this?... seems like changes here would potentially be breaking, I'd like to understand why that is a good thing / bad thing.

[iherman]

I find the concept of "VerifiableCredentialGraph" confusing, and note that it does not occur even once in the current TR.

https://github.com/search?q=repo%3Aw3c%2Fvc-data-model%20VerifiableCredentialGraph&type=code

Indeed, it is not explicitly in the spec. But the concept is already there in the context file:

"verifiableCredential": {
          "@id": "https://www.w3.org/2018/credentials#verifiableCredential",
          "@type": "@id",
          "@container": "@graph"
        },

Ie, the fact that the value of a verifiableCredential is a graph and not a VerifiableCredential. That aspect must be expressed in the vocabulary. Speaking bluntly, the vocabulary specification in the 1.1 version for that term was a bug.

Maybe it is worth referring to the concept of RDF graphs (a.k.a. datasets) in the RDF 1.1 spec.

I prefer to remove the term, since it can only lead to more confusion.

I would -1 to remove the "concept" from the vocabulary. I am fine trying to find a different name.

[iherman]

Admin note: this issue was discussed and decided upon a long time ago in issue #770, and a resulting PR in #771 (back in 2021!). I have made the mistake (due to my failing memory) to reopen the same issue a year later in #907 and the general concept was also discussed in #881. All these issues and PR-s have been closed, and the current vocabulary is around for at least a year. Is there a reason for reopening the same issue again?

[OR13]

Ie, the fact that the value of a verifiableCredential is a graph and not a VerifiableCredential. That aspect must be expressed in the vocabulary. Speaking bluntly, the vocabulary specification in the 1.1 version for that term was a bug.

Yes, I thought these were related.

I guess I am still confused on the difference between a "graph" and a "JSON-LD object with a context".

Especially given this:

• https://w3c.github.io/vc-data-model/#dfn-graphs
• https://w3c.github.io/vc-data-model/#dfn-credential

[OR13]

Admin note: this issue was discussed and decided upon a long time ago in issue #770, and a resulting PR in #771 (back in 2021!). I have made the mistake (due to my failing memory) to reopen the same issue a year later in #907 and the general concept was also discussed in #881. All these issues and PR-s have been closed, and the current vocabulary is around for at least a year. Is there a reason for reopening the same issue again?

Yes, in v2... both the context and vocabulary are now normative... and I suspect based on the issues linked that people still don't understand the implications of this on verifiers.

[iherman]

Admin note: this issue was discussed and decided upon a long time ago in issue #770, and a resulting PR in #771 (back in 2021!). I have made the mistake (due to my failing memory) to reopen the same issue a year later in #907 and the general concept was also discussed in #881. All these issues and PR-s have been closed, and the current vocabulary is around for at least a year. Is there a reason for reopening the same issue again?

Yes, in v2... both the context and vocabulary are now normative...

I do not think that declaring them to be normative makes a change in this respect. Actually, because the vocabulary is normative, this bug MUST be handled (and has been handled in the v2 version of the vocabulary), whereas v1 could be published with a buggy vocabulary.

[OR13]

@iherman we can make breaking changes in v2... fixing bugs should be on the table, not under the rug : )

It's possible the working group agrees, that the current behavior / bug is ok... its possible they want to fix it.

How do we "handle this bug", what does that mean?

[iherman]

I do not understand your remark, @OR13. This issue was raised, discussed, and closed. Long time ago.

At this moment, there is no bug. The vocabulary has been handled/fixed a while ago.

You are questioning the underlying concept (a.k.a. the fact that verifiableCredential is labeled as being a graph container in the context file) which has been known for a long time, and I do not see any new fact on the table that requires reopening this.

[OR13]

@iherman concretely, I would like to see this part of the vocabulary and the related parts of the TR improved... substantially:

- id: verifiableCredential
    defined_by: https://www.w3.org/TR/vc-data-model-2.0/#defn-verifiableCredential
    label: Verifiable credential
    domain: cred:VerifiablePresentation
    range: cred:VerifiableCredentialGraph
    comment: The value of this property identifies a <a href="#VerifiableCredentialGraph">Verifiable credential graph</a>. (Informally, it indirectly identifies a <a href="#VerifiableCredential">Verifiable credential</a> contained in a separate graph.)

The new information on the table is that I am asserting that it is confusing, and not good enough now that the vocabulary is normative.

1. the label should be updated.
2. the comment should refer to the TR, and explain the usefulness of "separate graphs".
3. ideally the range can be adjusted to not introduce a new concept... For example, would it be equally precise to say the range is "Credential" (related to the Abstract class issue with "VerifiableCredential") ?

In short, I don't think the current vocabulary meets the standard for publication, and especially this term, needs improvement.

Looks at the rest of the vocabulary, I can see several other issues... I will file some PRs to correct them.

[OR13]

https://github.com/w3c/vc-data-model/pull/1241/files

What is the utility of label ? It seems to not provide a lot of value in the current vocabulary... How can it be improved?

[iherman]

@iherman concretely, I would like to see this part of the vocabulary and the related parts of the TR improved... substantially:

- id: verifiableCredential
    defined_by: https://www.w3.org/TR/vc-data-model-2.0/#defn-verifiableCredential
    label: Verifiable credential
    domain: cred:VerifiablePresentation
    range: cred:VerifiableCredentialGraph
    comment: The value of this property identifies a <a href="#VerifiableCredentialGraph">Verifiable credential graph</a>. (Informally, it indirectly identifies a <a href="#VerifiableCredential">Verifiable credential</a> contained in a separate graph.)

The new information on the table is that I am asserting that it is confusing, and not good enough now that the vocabulary is normative.

1. the label should be updated.

I welcome any specific proposal.

2. the comment should refer to the TR, and explain the usefulness of "separate graphs".

I leave this to @msporny, @dlongley, @awoie or others who are deeply involved with the VC details on the matter. The comment field can refer to any specific text of course.

3. ideally the range can be adjusted to not introduce a new concept... For example, would it be equally precise to say the range is "Credential" (related to the Abstract class issue with "VerifiableCredential") ?

The "new concept", as you call it, is absolutely necessary, because an RDF Graph is not the same as a Class. Dropping this re-introduces the bug in v1. And, I repeat, it is not a "new concept" at all: it is a proper formulation of the fact that the verifiableCredential property (or the proof property in the security vocabulary, for that matter) are defined as graph containers. That has always been the case ever since I looked at the verifiable credentials specification. Removing this feature is a hugely breaking change.

Unless you define "Credential" as a Graph and not a Class, your proposal will not fly either.

cc @pchampin

[TallTed]

What is the utility of label?

It's a human-facing label for the machine-facing URI. Feed the ontology to a visual editor, and you should see the utility. Instead of a bunch of nodes and edges that are all visually labeled with URIs, you'll see them labeled with labels.

[OR13]

Perhaps we can elaborate on the value of "VerifiableCredentialGraph" in the processing the vc-data-model as JSON-LD section? cc @msporny

[OR13]

It's a human-facing label for the machine-facing URI. Feed the ontology to a visual editor, and you should see the utility. Instead of a bunch of nodes and edges that are all visually labeled with URIs, you'll see them labeled with labels.

thanks @TallTed makes sense that we would then strive to make labels more accurately describe the term URIs, so adding the word "graph" here... makes sense... https://github.com/w3c/vc-data-model/pull/1241/files#diff-dfe00db9d510a45f71289312990ee36df870f36c02094551bfba625dcd0d8e96R155