Re: A modest proposal concerning blank nodes.

Nathan wrote:
> Pat Hayes wrote:
>> On Mar 2, 2011, at 5:46 PM, Nathan wrote:
>>
>>> Nathan wrote:
>>>> mini proposal 2..
>>>> only one simple question, why not scope blank node identifiers to 
>>>> named g-boxes where a name is present, and where not as a syntax of 
>>>> the serialization (well, to the g-text), this allows people to 
>>>> migrate to blank nodes being persistent identifiers without breaking 
>>>> bc, and keeps that existential quantification in RDF.
>>> s/persistent identifiers/persistent local identifiers/
>>> as in, scoped inside a g-box, as a property of a g-box where one 
>>> exists, and where not as a property of a g-text (which is just an 
>>> anonymous g-box when you abstract enough).
>>
>> Yes. :-)
>> See my reply to Richard.
> 
> cool - just to confirm in a sub thread, this means that if a g-box with 
> a name exists (say just a normal ir on the web, or a FROM <graph> in 
> sparql) then two subsequent data retrievals would result in _:b1 being 
> _:b1 in both.
> 
> if however it was a quoted graph, or some anonymous chunk of RDF, then 
> the blank node identifier would not be maintained by default.
> 
> although, in the second case, and by way of niftyness, if it's scoped to 
> the g-box, then to parse or serialize a graph you have to "put it in a 
> box" so the names could be maintained (? perhaps)
> 
> the question then, when you take two g-boxes with names, and smash the 
> triples together, how do you prevent bnode-identifier conflicts? do the 
> names get replaced, or does this mean bnode-identifiers become a pair
>   ( <g-box-name> , _:b1 ) for named g-boxes
>   ( , _:b1 ) for anonymous g-texts
> 
> or?

<doh> it wouldn't be an issue, because when you smash only the triples 
together, they loose the associated name and become a new anonymous 
graph, it's a wilful stripping of the name, or provenance if you like.

Received on Thursday, 3 March 2011 00:58:57 UTC