Re: [Tracker] does/will Tracker expose a SPARQL endpoint?

[<ivan frade nokia com>]

Hi!

 It is possible to write an http endpoint that proxies tracker (Mark did it before). IIRC, he did it in 
python using tracker DBus interface. Now that we have libtracker-sparql, i wonder how difficult would it be 
to write an Apache module to act as http endpoint.

There are still few points you should take into account:

1. Tracker is a "compiled" triplet store: it needs to know the ontology in advance, and it will handle 
triplets ONLY in that ontology.
  This is not dramatical: you can start tracker with your own ontology without much problem, but well, makes 
it different from other generic stores.

2. Tracker is designed for a single user in a desktop session. (In other words, each user has its own 
tracker).
  You can trick it to use it as a server behind an endpoint with an specific user, but the security/access 
control is exactly that: what the user can read. 

 This is also related with your last comment. RDF + security restrictions is a tricky combination, and 
offering your desktop data without any control is not nice. Of course, any input of these issues is welcome!

 Regards,

Ivan
________________________________________
From: tracker-list-bounces gnome org [tracker-list-bounces gnome org] on behalf of ext Rob Taylor [rob taylor 
codethink co uk]
Sent: 06 January 2011 11:08
To: Adrien Bustany
Cc: tracker-list gnome org
Subject: Re: [Tracker] does/will Tracker expose a SPARQL endpoint?

CCing in Mark on this for his response.

Rob

On 06/01/11 08:26, Adrien Bustany wrote:
> Le Thu, 6 Jan 2011 01:57:18 -0600,
> Darren Cruse <darren cruse gmail com> a écrit :
>
> > Hi guys,
> >
> > Brand new to tracker apologize if this has already been asked but
> > didn't see it in the archives...
> >
> > I'm working on a project creating RDF metadata from a fairly large
> > pool of html documents where the results have to be searchable
> > remotely in what we'd assumed would be a web interface.
> >
> > It's funny a first stab at the system has been done using an approach
> > that looks a lot like tracker, if tracker had a SPARQL endpoint where
> > we could transform it's SPARQL xml response into html for the web
> > based search UI (in our case using our preferred toolset which for us
> > would be XQuery and/or java, not C/C++).
> >
> > But I don't see mention of a SPARQL endpoint for tracker?
> >
> > Is there one and/or is there plans for one?
> >
> > Or does using Tracker for this purpose feel at odds with how Tracker's
> > envisioned?
> >
> > (though there's mention of metadata "sharing" in some of the
> > literature I'm seeing on "Semantic Desktop" in my brief look at
> > Tracker and Nepomuk KDE I'm not clear how to get at the RDF metadata
> > on anything but the machine it's been indexed on)
> >
> > Thanks,
> >
> > Darren
> > _______________________________________________
> > tracker-list mailing list
> > tracker-list gnome org
> > http://mail.gnome.org/mailman/listinfo/tracker-list
>
> Hello Darren,
>
> I think this was done once by Mark Doffman (sorry if I'm misspelling),
> who was running some benchmarks against tracker. You might want to
> check with him. Doing one from scratch shouldn't be too hard, it's just
> a matter of running the query and formatting the results properly, all
> the heavy lifting is done by Tracker.
>
> Cheers
>
> Adrien
> _______________________________________________
> tracker-list mailing list
> tracker-list gnome org
> http://mail.gnome.org/mailman/listinfo/tracker-list


--

Rob Taylor, Managing Director, Codethink Ltd. - http://codethink.co.uk
Twitter: @robtaylor78 - LinkedIn: http://www.linkedin.com/in/robtaylor78
Office: +44 161 236 5575 - Cell: +44 7891 533856
_______________________________________________
tracker-list mailing list
tracker-list gnome org
http://mail.gnome.org/mailman/listinfo/tracker-list