- From: (unknown charset) Chimezie Ogbuji <ogbujic@bio.ri.ccf.org>
- Date: Mon, 29 Jan 2007 13:10:31 -0500 (EST)
- To: (unknown charset) Matthias Samwald <samwald@gmx.at>
- cc: (unknown charset) public-semweb-lifesci@w3.org
- Message-ID: <Pine.GSO.4.60.0701291254420.29468@joplin.bio.ri.ccf.org>
On Mon, 29 Jan 2007, Matthias Samwald wrote: > > >> Well, I'm inclined to ask why URI's should be explicitely assigned? > > We had a discussion about blank nodes on a teleconference some months ago: > http://esw.w3.org/topic/HCLSIG_BioRDF_Subgroup/Meetings/2006-09-11_Conference_Call > > bNodes don't serve as a replacement for URIs for cases where you cannot decide on assigning a URI. I guess for me, the question is what is the intent of the node itself? Are you *really* trying to make an existential statement and if you are why not use the semantics that RDF-MT is natively built to support instead of trying to workaround corner cases. > They behave somewhat different than one might think, e.g. in a SPARQL query. Not to open up this particular can of worms, but I consider (and this is just my opinion, mind you, I'm not trying to be derogatory with regards to the DAWG work in this regard) the treatment of 'told' BNodes in a query as a corner case which conflates how BNodes are meant to be used semantically with how nodes are 'handled' between query sessions. You can't on the one hand rely on the underlying expressivity of RDF-MT semantics and then avoid it because of a contentious query syntax usecase. Either RDF supports 'true' existential variables (in the sense in which the language it 'burrows' the semantics from meant them to be used: Predicate Logic) or it doesn't try to at all. > Another obvious reason for assigning URIs instead of bNodes is that URIs are essential for making connections between distributed datasets, which is one of the main arguments for using Semantic Web standards. Well then, the intent here isn't to make an existential statement, because if you rely on those identifiers being resolvable (and thus permanent) then you *do* need to 'hijack' the skolemization process. But we must take care to not make it a blanket statement to cover all usages of BNodes in a Semantic Web. Existential variables *do* have a semantic purpose and precedent (which predates the SW). The 'best practice' of having resolvable URIs should not be conflated with the value in being able to make minimal (but binding) statements about a 'thing'. Otherwise that argument could easily spill over into how existential restrictions are used in DL and require that every owl:Restriction is a 'named' class. > Even if you are sure that your ressource is of no interest for anyone besides yourself at the time you are creating it, this might still change later on. It's not so much a question of whether or not the resource is of interest, but of of whether the intent is to make an existential statement. If I say: 'someone loves me' I'm not *only* saying that I don't know the person's name but also that *at the very least* there is some such person. It's not a straight forward decision when to use BNodes, but the idea that you shouldn't because we can easily mint URIs bothers me. Chimezie Ogbuji Lead Systems Analyst Thoracic and Cardiovascular Surgery Cleveland Clinic Foundation 9500 Euclid Avenue/ W26 Cleveland, Ohio 44195 Office: (216)444-8593 ogbujic@ccf.org
Received on Monday, 29 January 2007 18:10:38 UTC