Re: DBpedia now available as triple pattern fragments

Dear all,

>> Kudos! Fantastic work Ruben. Is there already a public status page comparing
>> the uptime of DBpedia's SPARQL endpoint for a query equivalent to a TPF
>> request?

I use Pingdom to track the availability of the triple pattern fragments server.

I also track the DBpedia public SPARQL endpoint; however this is more tricky.
The Pingdom server only pings one URL, and I have sometimes get cached results,
even at times when DBpedia was down.
So it does not accurately reflect DBpedia's actual downtime;
tests like what SPARQLES has done seem to be better for that.

Results will be made available publicly,
as well as the monthly cost of hosting the server.

> The utility of that serves what purpose?  I think Ruben clearly described this as a complimentary addition.

Exactly _because_ I described them as a complementary addition.
It wouldn't make sense to host DBpedia as triple pattern fragments
if the DBpedia SPARQL endpoint had high availability.
After all, the DBpedia SPARQL endpoint solves SPARQL queries much faster
when it is available—which is not always the case.

The complementary nature of a SPARQL interface and a triple pattern fragments interface
is in the trade-off speed/availability.
Solving SPARQL queries over triple pattern fragments is at least a magnitude slower,
but we get higher availability.

So if people ask: “how are they complementary?”,
then this is the honest answer we need to give them.

> That said, if you feel there's value in comparing both, I'll be happy to invest some time in telling a complete story,  FWIW.

I can suggest my ISWC2014 slides (and talk, a recorded version will be published),
as they really emphasize the trade-offs involved in the different interfaces:
http://www.slideshare.net/RubenVerborgh/querying-datasets-on-the-web-with-high-availability

Best regards,

Ruben

Received on Friday, 31 October 2014 09:33:50 UTC