Re: defn of blank nodes and triples RE: Review of 21-Nov concepts draft, part 3, sect 3-end

>[[This is an offlist thread I am moving on-list because I think this is of
>wider interest]]
>
>Graham and I are discussing the definition of a triple.
>
>In the text of the current published draft:
>
>http://www.w3.org/TR/2002/WD-rdf-concepts-20021108/#xtocid103646
>
>which I essentially want to leave unchanged, the term blank node is not part
>of the defintion of a triple, instead blank node is defined later, depending
>on this definition of triple.
>
>It is, however, confusing to suppress the key concept of blank node to the
>extent that the current published draft does.

Yes, I think it is confusing. Reading that draft I immediately 
thought, so what can be a subject? Can my auntie Nance be a subject, 
or a sodium atom? Or is it supposed to be something syntactic??

>
>There is something quite elegant about the current negative definition of a
>blank node - what is a blank node? we don't tell you, we only tell you what
>it isn't. This elegance does capture something of the elusiveness of blank
>nodes.

I wish they weren't so elusive-seeming. They are a very simple idea, 
and they ought to seem very simple.

>
>Graham prefers text like:
>
>Graham:
>>  > An RDF triple contains three components:
>>  >
>>  > the subject, which is an RDF URI reference or a blank node
>>  > the predicate, which is an RDF URI reference
>>  > the object, which is an RDF URI reference, a literal or a blank node
>>  >
>
>This is reminiscient of Pat's  earlier texts. Pat made these work by having
>an explicit ready supply of blank nodes.
>My text has an implicit ready supply of nodes.

It might be too implicit, seems to me. We know what these things are: 
they are essentially bound variables, but we don't bother to give 
them a name. If someone wants to be sure that enough bnodes exist, we 
can say that they can use bnode identifiers as bnodes.

>
>Graham:
>>  Well, IIRC, you say later that a blank noide is one that isn't a URI or a
>>  literal.  I considered that the term blank node here would just be a
>>  forward reference to that description, as are URI reference and literal.
>>
>>  I don't think this is critical.
>
>My worry is that we then get to a circular defn:
>
>Currently:
>
>blank node depends on graph depends on triple depends on (uriref and
>literal).
>
>If blank node is added to the defn of triple we get:
>
>blank node depends on graph depends on triple depends on (uriref and literal
>and blank node).
>

I don't think it is circular to say that bnodes are some set of 
things that are not urirefs or literals and are called blank nodes, 
and that they can be in triples. We don't *define* them, because we 
don't need to. But we can say that there are denumerably many of them 
and that they are therefore in 1:1 correspondence to bnode IDs. So if 
anyone wants to use bnodeIDs for bnodes, they are free to do so.

Pat

-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam

Received on Sunday, 1 December 2002 00:17:34 UTC