W3C home > Mailing lists > Public > public-lod@w3.org > March 2009

Re: Finding SPAQL endpoints?

From: Kingsley Idehen <kidehen@openlinksw.com>
Date: Sat, 07 Mar 2009 16:16:17 -0500
Message-ID: <49B2E421.5040008@openlinksw.com>
To: Daniel Schwabe <dschwabe@inf.puc-rio.br>
CC: public-lod@w3.org
Daniel Schwabe wrote:
> Kingsley , Giovanny, Michael  and all,
> thanks for the prompt replies.
>
> I didn't express myself properly. I'm looking for a programmatic way 
> to find the SPARQL endpoint, something analogous to a DNS -> ip 
> translation. This metaphor breaks down right away, as there is no 
> centralized service where I can query - this is what I was trying to 
> emulate with Sindice, by issuing a query for the void:sparqlEndpoint 
> property, given the dataset's home  URL. Hopefully, if there is one, 
> Sindice would have indexed it.
In our case when you de-reference a URI we include an rdfs:seeAlso link 
to an associated VoiD graph URI. The idea being you discover the 
endpoint as part of the resource description.
>
> I could query the site for its sitemap extension (would it always be 
> <home url>/sitemap.xml? doesn't seem so...), as Giovanni suggests, and 
> see if I get a result; in the affirmative case, I have to parse it and 
> look for the <sc:sparqlEndpointLocation> element.
> As Michael Lang put it, one could have a Dataset Registry, but again I 
> would want to have a way to query it programmatically. In addition, it 
> would be a centralized resource.
>
> Hence my suggestion - I am looking for a simple way to enable 
> autodiscovery. If all sites would agree to make the sitemap.xml 
> available at <home url>/sitemap.xml, that would work as well. I just 
> think that a simple naming convention such as <home url>/sparql would 
> be even easier to implement, and it may become one of those "de facto" 
> standards if the main datasets follow the convention... (as I said, 
> many already do...)
I also see utility in complementing what I stated above with  the following:

1. "sitemap.xml" in: http://<cname>/sparql/

2.  Link hook from <head/> section of  http://<cname>/sparql as in:

 <link rel="rdfs:seeAlso" type="<content-type>" title="Data Set 
Description (VoiD Graph)" 
href="<void-graph-info-resource-interface-url>" />

3. A de-referencable URI for VoiD graphs (we just need to agree a URI 
pattern for Slash or Hash schemes)


Kingsley

>
> Cheers
> D
>
>


-- 


Regards,

Kingsley Idehen	      Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO 
OpenLink Software     Web: http://www.openlinksw.com
Received on Saturday, 7 March 2009 21:16:53 UTC

This archive was generated by hypermail 2.3.1 : Sunday, 31 March 2013 14:24:19 UTC