Is https:// a valid IRI according to RFC 3986?

[bnouwt]

Hi everyone,

I am updating our software to use apache jena 6.0.0 and are therefore using jena-iri3986 to check validity of IRIs. One of our unit tests started failing after these changes, specifically because https:// was expected to not be a valid absolute IRI, but IRI3986.create("https://").isAbsolute() returns true. The isAbsolute() method refers to RFC 3986 for the definition of an absolute IRI, but I have difficulty understanding whether the authority is required or not. Can someone explain?

Thanks in advance!

Barry

[afs]

Hi Barry,

Which version are you upgrading from?

You should be able to use IRIx and not need to use IRI3986 as of Jena 6.0.0

It is "absolute" - RFC 3986 URI syntax allows the hier-part to have a host that is an empty string for reg-name. "absolute does not mean much more than "has scheme, does not have fragment".

However, that's just the basic syntax. Mistakes in URI syntax are exceptions.

URI schemes are layered on top of the URI syntax; errors and warning in URI schemes are recorded with the IRIx.

https/http (also defined in RFC 3986) require a non-empty host - that's https/http specific.

Jena will parse that and print a warning for errors in scheme-specific rules. Parser errors are fatal errors.

        IRIx iri = IRIx.create("https://");
        iri.handleViolations((err,str)->System.out.printf("Error=%s : %s\n",(err?"yes":"no"),str));
        System.out.println(iri.isReference());

prints

Error=yes : <https://> http and https URI schemes do not allow the host to be empty
true

the "isReference" value is a Jena bug (#3825). It should be false.

RDF 1.2 Concepts introduces RDF Reference IRIs" in section 3.3.1.
These are advisory - and includes the advice "follow the URI scheme rules". IRIX.isReference needs to be brought up-to-date.

[bnouwt]

Hi Andy, thanks for the swift reply. We are upgrading from Apache Jena version 5.6.0 to Apache Jena 6.0.0.

Your response triggered us to rethink what we were actually trying to achieve with our test and this caused us to realize we need valid RDF IRIs. Via your IRIx reference we found IRIs.check(...) which seems to work as hoped.👍

Best regards,

Barry