- From: Ruben Verborgh <ruben.verborgh@ugent.be>
- Date: Fri, 31 Oct 2014 10:46:12 +0100
- To: Kingsley Idehen <kidehen@openlinksw.com>
- Cc: public-hydra@w3.org
Hi Kingsley, > And the uptime means what? It makes all the difference whether or not you can build a reliable application on top of something. > we can begin to understand and appreciate the complimentary nature of LD Fragments in regards to Linked Open Data deployment. That's what I'm advocating and always will do: http://www.slideshare.net/RubenVerborgh/querying-datasets-on-the-web-with-high-availability/43/ Summarized for consumers of Linked Data: - Want fast, live SPARQL queries? Use a public SPARQL endpoint. - Want fast SPARQL queries with high availability? Use the data dump and set up a private SPARQL endpoint. - Want live SPARQL queries with high availability? Use the triple pattern fragments interface. Summarized for publishers of Linked Data: - If you have budget for a public SPARQL endpoint, set up a public SPARQL endpoint. - If you don't have budget for a public SPARQL endpoint, set up a triple pattern fragments interface. - In any case, always offer a data dump, so people can do what they want. In the future, other combinations of trade-offs will exist; for now, we don't have “fast + live + high-availability + low-cost” yet. > Creating the illusion of 99.99% uptime isn't an issue here It's not an illusion; fragments.dbpedia.org has 100% uptime so far, as measured by Pingdom, an independent party. Statistics will be published. But this 100% uptime comes with a cost: queries are slower; some queries are very slow. The public DBpedia SPARQL endpoint gives you fast queries; the price you pay is availability. We cannot be more honest than that. Best, Ruben
Received on Friday, 31 October 2014 09:46:45 UTC