Re: RDF's relative IRI resolution is ambiguous

Hi Richard,

> IRI normalisation may very well be performed if the syntax specification requires it, as part of relative IRI resolution or otherwise.

Indeed; however, the problem I'm bringing up is that relative IRI resolution is not well-defined in those syntaxes,
because the referenced RFC for resolution performs normalization.
And given the following

> The RDF Concepts spec only states that when *testing for IRI equality in an RDF graph*, IRIs are not normalised.

we cannot perform IRI normalization in general
because normalization of non-normalized IRIs lead to a non-equivalent graph.

With an example: if a Turtle document contains

    BASE <http://example.org/xxx/yyy/zzz/../../../>
    <a> <b> <../aaa/bbb/ccc>.

the Turtle syntax spec does not conclusively say how to interpret this triple,
and we cannot simply normalize all of its components to make the interpretation,
because IRI normalization under the RDF model leads to a different graph.

Best,

Ruben

Received on Thursday, 27 August 2015 14:08:20 UTC