Re: saddle?

Eric Prud'hommeaux wrote:
> On Tue, May 16, 2006 at 05:00:52PM +0100, Seaborne, Andy wrote:
>>
>>
>> Kendall Clark wrote:
>>> Folks,
>>>
>>> I've had two separate queries in the past few days about SADDLE. So, I 
>>> have two questions:
>>>
>>> 1. does anyone have a copy of any of that work? I fear it only exists in 
>>> CVS somewhere, as an old revision of the editor's draft of the protocol 
>>> spec. I'll be trawling through that looking for bits, but if someone has 
>>> something else ready-to-hand, that would be great.
>>>
>>> As I recall the majority of the work done on this was done at the Boston 
>>> F2F. Maybe someone has some notes from that?
>>>
>>> 2. is there any interest in restarting work on SADDLE sooner rather than 
>>> later, on the assumption that 2 recent queries probably may be the tip 
>>> of an iceberg -- okay, probably not iceberg, but if 2 people asked, I 
>>> assume lots of other people wondered and didn't ask.
>>>
>>> Cheers,
>>> Kendall
>> A couple of data points:
>>
>> http://esw.w3.org/topic/SparqlEndpointDescription
>>
>> and there is a student (Bastian Quilitz) here working on the problem from 
>> the point of view of information needed to federate queries.  That work 
> 
> indulgent question:
> 
> Does that involve some sort of Query Pattern patterns?

yes - the splitting of the query patterns needs to know the specificity of the 
patterns to do it in the right order.

The demo Bastian has queries across the enterprise LDAP server (presented as 
SPARQL queries via Damian's code) and the HPLabs technical report library.

Q1: "who works with the author of report entitled 'ABCD'?",

then should go to the technical reports DB first but

Q2: "what reports have people who work with "Fred" written in the last 2 years?"

then should go to the LDAP server first


>  If I go to
> IMDB to resolve queries with certain predicates:,
> 
> { ?m imdb:name ?movieName .
>   ?m imdb:star ?starName }
> 
> they will only answer if one of those is constrained:.
> 
>   [
>     wsdl:endpoint <imdb> ;
>     saddle:answers [
> 	saddle:pattern "{ ?m imdb:name ?movieName .
>                           ?m imdb:star ?starName }" ;
>         saddle:mustBeConstrained (?movieName) ] ;
>     saddle:answers [
> 	saddle:pattern "{ ?m imdb:name ?movieName .
>                           ?m imdb:star ?starName }" ;
>         saddle:mustBeConstrained (?starName) ] ;
>   ]

It's more than just patterns - it's selectivity (weighted patterns).

> 
> I've been curious about this for a bit so I'm wondering if Basian is
> solving this sort of problem. (Also, I had to see if my mail probs
> were fixed, and this mail is pretty non-critical.)

Email seems to work!

	Andy

> 
>> looks promising although it's evolving nature may also be a sign that the 
>> problem space is not completely scoped out yet.  Problem definition woudl 
>> seem to be a possible next step (XG? IG? ad hoc on the ESW wiki?)
>>
>> 	Andy
>>

Received on Sunday, 28 May 2006 17:20:19 UTC