The WebID could not be parsed

[DylanVanAssche]

My webID https://dylanvanassche.be/#me is written in JSON-LD as snippets in my HTML website, no content-negotiation.
It seems that the CSS has issues to parse this. Would be useful if it would read the JSON-LD instead of forcing Turtle: https://github.com/CommunitySolidServer/access-token-verifier/blob/main/src/algorithm/retrieveWebidTrustedOidcIssuers.ts#L13=

CSS logs:

1|CSS inst | 2022-06-10T09:10:58.518Z [BaseHttpServerFactory] info: Received GET request for /idp/consent/
1|CSS inst | 2022-06-10T09:10:58.907Z [BaseHttpServerFactory] info: Received GET request for /idp/consent/
1|CSS inst | 2022-06-10T09:10:59.683Z [BaseHttpServerFactory] info: Received POST request for /idp/consent/
1|CSS inst | 2022-06-10T09:10:59.780Z [BaseHttpServerFactory] info: Received GET request for /.oidc/auth/omMLMhg0SUujFBMinpZyh
1|CSS inst | 2022-06-10T09:11:00.393Z [BaseHttpServerFactory] info: Received GET request for /.well-known/openid-configuration
1|CSS inst | 2022-06-10T09:11:00.467Z [BaseHttpServerFactory] info: Received OPTIONS request for /.oidc/token
1|CSS inst | 2022-06-10T09:11:00.537Z [BaseHttpServerFactory] info: Received POST request for /.oidc/token
1|CSS inst | 2022-06-10T09:11:00.651Z [BaseHttpServerFactory] info: Received GET request for /.oidc/jwks
1|CSS inst | 2022-06-10T09:11:13.383Z [BaseHttpServerFactory] info: Received OPTIONS request for /availability
1|CSS inst | 2022-06-10T09:11:13.645Z [BaseHttpServerFactory] info: Received GET request for /availability
1|CSS inst | 2022-06-10T09:11:14.034Z [DPoPWebIdExtractor] warn: Error verifying WebID via DPoP-bound access token: The WebID could not be parsed.
1|CSS inst | Actual: Invalid RDF
1|CSS inst | Expected: A valid turtle document
1|CSS inst | 2022-06-10T09:11:14.041Z [WebAclReader] info: Reading ACL statements from https://dyvassch.pod.knows.idlab.ugent.be/availability.acl
1|CSS inst | 2022-06-10T09:11:14.044Z [PermissionBasedAuthorizer] warn: Unauthenticated agent has no read permissions

[matthieubosquet]

A WebID document MUST be available as turtle as per the spec: https://www.w3.org/2005/Incubator/webid/spec/identity/#introduction.

I could argue that it makes sense if anything because having one base format is good for interoperability. Json-ld also happens to be problematic because all contexts must be dereferenced in order for it to be parsed as rdf.

[DylanVanAssche]

That's sad, that means that websites deployed to something like Github/Gitlab pages are all broken then as they do not support content negotiation.

[joachimvh]

A WebID document MUST be available as turtle as per the spec: https://www.w3.org/2005/Incubator/webid/spec/identity/#introduction.

The Solid spec only mentions that it should be an RDF document though so should perhaps be updated there for clarity? https://solidproject.org/TR/protocol#webid

Similarly for the OIDC spec https://solid.github.io/solid-oidc/#concepts-webids

(I tried looking for this requirement but couldn't find it is why 😄 ).

[matthieubosquet]

Yes, it’s a bit unfortunate that GitHub pages don’t support content negotiation.

I guess we could look into adding json-ld and rdfa parsing to the library.

[DylanVanAssche]

I plan to move my website at some point to my own hosting which I fully control, but I haven't been able to do it yet.
However, that would solve my specific case, but I can imagine a lot of people use these services for serving their website...

[joachimvh]

I guess we could look into adding json-ld and rdfa parsing to the library.

I'm not saying that you should since you're right that the spec requires it, but if you do you could just plug in rdf-dereference. Although that brings back the Comunica dependencies.

[elf-pavlik]

This lib might be used by clients as well #211 , including PWA. I don't think we should throw a bunch of RDF parsers in.

My webID https://dylanvanassche.be/#me is written in JSON-LD as snippets in my HTML website

Turtle can be embedded in HTML in a similar way https://www.w3.org/TR/turtle/#in-html

I think it shouldn't be that hard to pick out the <script type="text/turtle"> tag and parse the turtle in it. It shouldn't require any additional parsers at least when run in a web browser.

[rubensworks]

Just want to chime in here to signal that the strict Turtle requirements on WebIDs is under discussion here: w3c/WebID#3

My personal view on this is that WebIDs should be readable in any RDF serialization.

As @joachimvh has said, parsing from all RDF serializations can easily be supported in clients using libs such as rdf-parse and rdf-dereference.