Re: Link relation type to link to discover LDP

>> I don't see technical problems using blank nodes as identifiers.
>> We will need to explicitly mention this in ExplicitRepresentation though.
> 
> What exactly needs to be mentioned?

How to encode blank nodes (i.e., the underscore-colon notation).

> Btw. when was this feature introduced? Somehow I completely missed that. I
> still had the max. 8 static views on a given triple (fixed value/arbitrary
> value for each triple component) in mind which isn't true with this anymore.

It was never too explicitly mentioned.
Mostly, we said 8 views, there was a footnote clarifying
that this number disregards patterns where one variable is used multiple times.

I guess it never really came up because queries that use such patterns are rare.
But they are real queries nonetheless.

> How do your clients deal with servers that return blank nodes nevertheless?

They can't ask for more patterns about blank subjects or objects.
Blank node identifiers only have meaning within one document;
so there is no way to refer to them later on.

So should a server return a data triple
    _:x a foaf:Person
then clients are unable to ask for more triples about this person.

At the moment, the spec says [1]:
    Triples SHOULD NOT contain blank nodes
    (if the original triples contain blank nodes, they SHOULD be skolemized).
but I'm actually thinking to make that a MUST NOT / MUST.

>> Does anybody see issues with using blank nodes to query patterns such as
>>   _:x foaf:knows _:y
> 
> I'm fine with it but I'm wondering whether this should still be called TPF
> or whether it should be a separate LDF interface.

I would still call this TPF, yes.
It's so closely related to everything else in LDF that a separate interface would not be meaningful.
It would also be very hard to explain in hypermedia to a client the difference between with and without.

Furthermore, if the server does not support it, there's no problem:
they will just return too much data, so that doesn't endanger completeness.

Best,

Ruben

[1] http://www.hydra-cg.com/spec/latest/triple-pattern-fragments/#data

Received on Sunday, 28 June 2015 13:01:03 UTC