GH-135: Follow URI scheme rules

[afs]

This closes #135.

This PR is close to the changes in #131 and #132 and will probably need adjusting after they are merged.

---

Preview | Diff

[pchampin]

As suggested in my review,
I would suggest to rephrase this as follows
(sorry, I can not use the suggestion mode, because the first line is not part of the diff)

IRIs in the RDF abstract syntax MAY NOT be relative references, and MUST be resolved per [RFC3986]. They MAY contain a fragment identifier.

I like the addition about following the rules defined by the scheme.

[afs]

Added. MAY NOT is not in RFC 2119; "MAY" means optional. I think you meant a "MUST NOT" strength condition.

[gkellogg]

I’m a bit concerned about saying that special rules without listing other important schemes, as this will lead to interoperability issues.

[afs]

I’m a bit concerned about saying that special rules without listing other important schemes, as this will lead to interoperability issues.

Which schemes did you have in mind? I picked HTTP (important) and DID (W3C) for the note.

URNs are complicated (there are rules per namespace).

file:///, even with RFC8089, are not global.

Some schemes don't exist! uuid:, oid: (properly, use the urn:uuid:, urn:oid: forms).

[gkellogg]

I’m a bit concerned about saying that special rules without listing other important schemes, as this will lead to interoperability issues.

Which schemes did you have in mind? I picked HTTP (important) and DID (W3C) for the note.

those certainly make sense given the context, but we make a broad statement that could include obscure schemes, or those yet to be released. White-listing specific schemes (http, https, and did, roof example) would enhance interoperability.

[afs]

White-listing specific schemes

Is the note text enough?
Or maybe move it up the list of points in the note?
Or change

IRIs SHOULD follow rules defined by the IRI scheme, such as those for HTTP/HTTPS [RFC7230].

URI Schemes: Examples of scheme-specific rules include the scheme rules for HTTP and the DID syntax.

I have changed it to "HTTP/HTTPS"

An advice paragraph somewhere to use specific schemes might be nice (but outside this PR).

Aside: ReSpec has used not used the more usual RFC reference -- https://datatracker.ietf.org/doc/html/rfc7230.

[pchampin]

I think @gkellogg 's concern could be addressed in the note, by rephrasing it:

URI Schemes: Implementations may ignore URI schemes that they do not recognized, but are encouraged to enforce the scheme-specific rules of the most common ones, such as scheme rules for HTTP/HTTPS and the DID syntax.

[gkellogg]

I think @pchampin’s suggested wording is okay. The SHOULD makes testing a bit challenging, though. Typically, we do not test SHOULD/MAY behavior.

[afs]

Note updated.

If they don't recognize a scheme, it isn't an option to ignore - they must!

For the testing, my ideal is having rdf-tests as a record of community practice with an emphasis on common practice to warn about.

"SHOULD" is there because (1) this is new, not in RDF 1.1 (2) the "rules" aren't always cut-and-dried.

[afs]

I will squash the PR to one commit before merge; for now, the PR is keeping each change as a commit.