Re: Can we afford to offer SPARQL endpoints when we are successful? (Was "linked data hosted somewhere")

2008/11/27 Richard Cyganiak <richard@cyganiak.de>

>
> Hugh,
>
> Here's what I think we will see in the area of RDF publishing in a few
> years:
>
> - those query capabilities are described in RDF and hence can be invoked by
> tools such as SQUIN/SemWebClient to answer certain queries efficiently
>

I still don't understand what SQUIN etc have that goes above
Jena/Sesame/SemWeb etc which can do this URI resolution with very little
programming knowledge in custom applications.

- everyone who wants more advanced query capabilities, will crawl the site
> and run their own local SPARQL store


I like the EC2 approach that Kingsley has been talking about precisely
because it stops people having to crawl unnecessarily. There is a huge
efficiency gain if you can just flick a switch inside Amazon (or other cloud
provider) and have a unique data and hardware instance created for you that
you can immediately have fully private SPARQL access to.

At the moment we don't have the technology for describing non-SPARQL query
> interfaces in RDF, and crawling linked data is still a fairly complex
> business. As long as these problems are not solved, we pretty much are stuck
> with SPARQL endpoints.
>

I am not sure what you mean by non-SPARQL query interfaces. The templating
work I developed that is currently implemented for Bio2RDF and DBpedia to me
shows that it is fairly simple to have generic URI based SPARQL query
interfaces, even if my implementation is only student grade quality right
now. You don't have to expose SPARQL endpoints to have other people use your
queries in this way.

Queries like the following just won't be possible if people lock everything
down to private access without opening up at least some standard query URI
interfaces. I have changed http://bio2rdf.org/ to
http://bio2rdf.mquter.qut.edu.au in these examples as this is my test server
with the software able to answer these queries.

http://bio2rdf.mquter.qut.edu.au/linksns/dbpedia/dbpedia:Sydney (Ie, all of
the reverse links in dbpedia to the resource denoted in bio2rdf URI notation
as dbpedia:Sydney, this query is actually answered by
dbpedia.org/sparqlcurrently btw although it could be easily changed to
point to a personal EC2
instance when they are available)

http://bio2rdf.mquter.qut.edu.au/links/go:0004396 (Ie, all of the known
resources that have been annotated with the Gene Ontology term named
"hexokinase activity")

Being able to query like this, and do more complex queries with a standard
vocabulary is a very interesting enterprise that requires some kind of
SPARQL backend at some stage.

Cheers,

Peter