Re: Re 2: PROPOSAL to close ISSUE-37: Clarifying bnode explanation

Hi Toby,

Sorry, but I think this expresses the same confusion that I've been
trying to clarify (comments inline).

On Tue, Oct 19, 2010 at 9:38 PM, Toby Inkster <> wrote:
> How about:
> """
> After processing, the following triples will be generated:
>  _:john foaf:mbox <> .
>  _:sue foaf:mbox <> .
>  _:john foaf:knows _:sue .
> The blank node identifiers ("_:john" and "_:sue") are arbitrary...

The "blank node identifiers" are *not* arbitrary; they are the
identifiers that were chosen by the author when creating the RDFa, and
they will drive the triples that are generated.

It does become arbitrary when you attempt to *retrieve* the triples
back from a triple-store; when the triple-store needs to serialise a
reference to a bnode it will have to generate fresh bnode identifiers
-- and they *will* be arbitrary.

> ...and implementations are not required to maintain the same identifiers as
> occur in the RDFa markup.

Implementations are not required to maintain bnode identifiers at
all...bnode identifiers exist at the level of the serialisation, not
in the store.

> The above data could have equivalently been
> represented as:
>  _:a foaf:mbox <> .
>  _:b foaf:mbox <> .
>  _:a foaf:knows _:b .

Not based on the example, it couldn't.

The bnode identifiers need to be the same *on the way in*. Your
example is fine when it comes to querying the triples *on the way
out*, but that isn't what the examples are trying to show. I.e., they
are not saying 'this is what you would get if you queried a
triple-store', they are saying 'this is what will be passed to a
triple-store for storage' (and by implication bnode identifiers will
be mapped to some internal representation).

> For clarity, this document retains blank node identifiers in examples...

I think this is also confusing; I'm not sure how else we would write
our examples if we didn't "retain" the identifiers?

> ...but developers must not rely on RDFa implementations returning
> identifiers that are consistent with the RDFa markup.
> """

I do think there are better ways of explaining this.



Mark Birbeck, webBackplane

webBackplane is a trading name of Backplane Ltd. (company number
05972288, registered office: 2nd Floor, 69/85 Tabernacle Street,
London, EC2A 4RR)

Received on Wednesday, 20 October 2010 08:36:33 UTC