- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Thu, 31 Jan 2002 11:49:55 +0200
- To: ext Sergey Melnik <melnik@db.stanford.edu>, Brian McBride <bwm@hplb.hpl.hp.com>
- CC: RDF Core <w3c-rdfcore-wg@w3.org>
On 2002-01-30 22:33, "ext Sergey Melnik" <melnik@db.stanford.edu> wrote: > From my perspective, the performace issue remains on the table as one of > the show stoppers for TDL (of course, it's a non-issue for toy data > sets, but may make million-triples data sets impractical). Let me offer only one possible implementation solution for TDL that is highly efficient: Transform the untidy TDL graph into a graph where every TDL pairing is expressed by a 'tdl:' URV. For literals that have no determinable type, use a consistent 'uuid:' URI for the datatype in the 'tdl:' URV. C.f. http://ietf.org/internet-drafts/draft-pstickler-tdl-00.txt E.g. OldTree ex:age _:1 . _:1 rdf:value "1984" . _:1 rdf:type xsd:integer . FavoriteBook dc:title "1984" . dc:title rdfs:range xsd:string . becomes OldTree ex:age <tdl:(xsd:integer)1984> . FavoriteBook dc:title <tdl:(xsd:string)1984> . [please forgive qnames in tdl: examples, they should be complete URIs...] For each input query, perform the same transformation. Now you have a query graph that is highly compressed, completely tidy for both lexical and URIref nodes, and provides value-based matching reliably and efficiently. Note that, because typed data literals are denoted now by URIs, all equivalent TDLs will be merged to be tidy. That doesn't mean that there will be a 1:1 correlation between tdl: URVs and values, since lexical forms can still be non-canonical -- but such a 1:1 correlation could be achieved by, as part of the transformation, converting all non-canonical lexical forms to canonical lexical forms (presuming all datatypes are known/supported by the transformation application). If you wish to also provide queries simply based on string equality of literals, then simply examine the local, literal portion of the 'tdl:' URVs during comparisons. No bloat. No increased inefficiency. No problem. As it has been said before, there are ways to achieve highly efficient implementations based on TDL. Cheers, Patrick -- Patrick Stickler Phone: +358 50 483 9453 Senior Research Scientist Fax: +358 7180 35409 Nokia Research Center Email: patrick.stickler@nokia.com
Received on Thursday, 31 January 2002 04:48:50 UTC