- From: Seaborne, Andy <andy.seaborne@hp.com>
- Date: Tue, 31 Aug 2004 17:57:27 +0100
- To: RDF Data Access Working Group <public-rdf-dawg@w3.org>
-------- Original Message -------- > From: Dirk-Willem van Gulik <> > Date: 31 August 2004 14:20 > > On Tue, 31 Aug 2004, Kendall Clark wrote: > > > Sorry, I don't mean to be nasty, the asemantics guys are great!, but > > these are all much worse than "?foo". > > Pick another one then :) - pretty much all other chars are safe except > for > the surrounders, _ and % if you ignore languages and just consider DB > interfaces. > > How about > > select =id, =title > where (=id ... variabnles appear elsewhere so this means =x == =y for ?x == ?y Parsers have no problems with this stuff - people do! > > to signal the 'link' ? OR how about just dropping it completely > > select id title > from <dbms://localhost/rfid> > where (id foo:id <urn:froefroe... > .. > using foo ... > > by simply insisting that any and all literals have a "" or '' , URI's > always have a <> around them and anything else must use a : separator > for the namepsace ? May work but others have objected to grammars that have lookahead >1. This has - need to look over the NCNAME to see if there is a :. x :p is var-x, followed by :p with the default namespace x:p is a qname. Confusing or what? And quds make it worse as a triple pattern + a space can become a valid quad pattern. > > Also think of the upside; do this right and you can instantly allow ANY > language/installation which has an ODBC, JDBC, ADO, whatever SQL > interface > do semantic web queries with virtually no rework on the client side. AND > you get at least some instant buy in from the people currently in > control > of data storage and the SQL bastion. We aren't querying existing SQL databases. > > Dw
Received on Tuesday, 31 August 2004 16:58:10 UTC