- From: Jos De_Roo <jos.deroo@agfa.com>
- Date: Sat, 15 Feb 2003 02:40:40 +0100
- To: Peter.Crowther@networkinference.com
- Cc: "WebOnt WG" <www-webont-wg@w3.org>, www-webont-wg-request@w3.org
>> From: Sean Bechhofer >> But what happens when I've got an ontology >> with 10^8 concepts/individuals in it and I want to do some simple >> processing on it, that doesn't necessarily warrant me >> building the whole data structure? > >As I mentioned at the Manchester f2f, this is exactly the problem we >have at NI. In our case, the simple processing is obtaining the OWL >statements, as distinct from the RDF statements, so that we can load >them into a reasoner. A 70,000 concept file takes 1.5G of RAM when >loaded into Jena (1.6), about 5-10 times the size of our internal >structures. We don't want to use a RDB simply to store the triples >during the load, as this would slow us down by orders of magnitude. >This particular customer wanted to give us OWL/RDF dumps with multiple >tens of millions of concepts, exported in arbitrary order. We have a testcase with a blob component that can deliver tens of millions of ***triples*** and I don't experience any scalabilty problems with that, to the contrary, it's much easier/faster to have distributed graph control. >We can stream OWL/XML, but not OWL/RDF, due to this annoying problem >that the triples describing a single concept nested to an arbitrary >depth may appear in an arbitrary order. What kind of nesting are you talking about? > We cannot find any way round >this given the (intentional) limitations of RDF. Unfortunately, Sean, I >think you're out of luck. -- , Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/
Received on Friday, 14 February 2003 20:41:18 UTC