Re: Moving forward with ISSUE-30 (IRI template expansion)

On Aug 19, 2014, at 3:42 AM, Ruben Verborgh <ruben.verborgh@ugent.be> wrote:

> Hi Markus,
> 
>> No, but you were actively involved in these discussions and provided a lot
>> of very valuable thoughts. So I want to make sure to get your feedback
>> before sending out another call for consensus.
> 
> Shout out to you for doing a great job listening to everybody here!
> 
>> Yeah, that's true. That's also what worries me about this approach but
>> apparently that's what the majority of the group wants. So my idea was to
>> explicitly mention that it is *simplified* Turtle (and thus not standard
>> Turtle) and explicitly call out the differences to the Turtle spec.
> 
> The strange thing about "simplified Turtle" is that
> it is *incompatible* with “full Turtle”, which is unexpected.
> Sure, “simplified Turtle” parsers would not be able to parse “full Turtle”,
> but the other way around is non-intuitive. The name is therefore inappropriate.

+1. The only "Simplified" Turtle I'm aware of is the variety used in Freebase dumps, which has a very constrained format. My thought is that we should be using specific constructs from Turtle, where it makes sense (i.e., literal encodings).

> The proper term would actually be
> “non-escaped N-Triples literal syntax with bracketless IRIs” 

Yes, but I wonder what the advantage of leaving brackets out for typed literals is; I can see it for non-literal values. As long as we can unambigiously distinguish between IRIs, plain literals, typed literals, and language-tagged literals. It's in the literal encoding that I think Turtle compatibility is most useful.

> “The corresponding RDF lexical form is the characters between the delimiters,
> <del>after processing any escape sequences</del>.
> If present, the language tag is preceded by a '@'.
> If there is no language tag, there may be a datatype IRI, preceded by '^^'.” [1]
> 
> I would just maybe non-normatively refer in the spec that the meaning of '@' and '^^'
> has been borrowed from Turtle/N-Triples, but call it something else.
> 
> Bear in mind that this could also be very confusing to readers:
> “Do my parameters have to be simplified Turtle is my document is JSON-LD?”
> 
>>> - TypedRepresentation (because we distinguish between literals and URIs)
>> 
>> I don't like this as much as I fear people will have a quick look, recognize
>> it as Turtle and move on.
> 
> Even worse with “simplified Turtle”.
> 
>> I personally feel better to explicitly acknowledge
>> that it is *based* on Turtle but not truly Turtle. Does this makes sense to you?
> 
> Not too much. The only thing it borrows is the meaning of '@' and '^^',
> there is no other relationship whatsoever.

Use Turtle literal-syntax only (other than the lack of prefix support).

Gregg

> Best,
> 
> Ruben
> 
> [1] http://www.w3.org/TR/n-triples/

Received on Tuesday, 19 August 2014 16:59:10 UTC