The productions for PN_CHARS_U are different between Turtle and N-Triples - this is visible in blank node labels.

N-triples is supposed to be subset of Turtle - it appears that it is N-triples that is wrong.

Turtle:

[164s] 	PN_CHARS_U 	::=   PN_CHARS_BASE | '_'

N-Triples:

[158s] PN_CHARS_U        ::=  PN_CHARS_BASE | '_' | ':'

This PR adds tests to turtle/ and ntriples/ assuming a correction for N-Triples.

I'll add copies to NQ and TriG.

It is looking increasing certain that NT and NQ are wrong and the intention was for there not to be : in the blank node label itself.

Let's wait to see if any parsers systems respond.

Let's wait to see if any parsers systems respond.

we accept it as n-quads, as that is what the syntax says.
we do not accept it as sparql, as that is what the syntax says.

The "Rio" Rust implementation passes all these tests without change (we applied the Turtle version of the rule).

Any reason not to merge this?

cc @lisp

Any reason not to merge this?

in itself no.
i do admit, however, that we continue not to recognize this as a failure.
i note also, that the tool to which we refer clients as a neutral reference for syntax issues - rapper, is also not sensitive to the restriction.

@afs, I'd say merge at your discretion.

See also https://www.w3.org/2001/sw/wiki/RDF1.1_Errata#erratum_30 with email discussions link.