W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > May 2005

RE: User defined operations and PREMISES keyword

From: Stephane Fellah <stephanef@imagemattersllc.com>
Date: Fri, 13 May 2005 14:28:30 -0400
To: <public-rdf-dawg-comments@w3.org>
Message-ID: <E1DWetw-00047q-Rq@lisa.w3.org>

Dan,

Thanks for answering my comments.

The FROM approach is valid only if the client has the ability to serve the
premises from a URI. This requires from the server the ability to resolve
the URI, establish the connection to a server and fetch the data using a
specific protocol. On client side, this would require to store the data on a
an external server (web server ?). This approach requires a lot of
infrastructure to put in place.

There are many case when clients generates premises in memory and want to
communicate directly with local query engine (ARQ for example and not from
the web).  For example, a client may draw a polygon on a map and asks the KB
server to provide all the Point of Interest within this polygon. 
The SPARQL language should provide a mechanism to pass this polygon inline
in the query. My proposal is addressing this need. 

I think the protocol should support both remote reference and inline
premises. 

>One implementation strategy that is consistent with the specification
>is for the engine to index its dataset before receiving queries and
>not handle on-the-fly additional premises.

I do not see how this could be accomplish, if the interactive client
generates premises on the fly (like the polygon drawn on the map).


Best regards

Stephane Fellah
Image Matters LLC


-----Original Message-----
From: Dan Connolly [mailto:connolly@w3.org] 
Sent: Thursday, May 12, 2005 10:22 AM
To: Stephane Fellah; public-rdf-dawg-comments@w3.org
Subject: Re: User defined operations and PREMISES keyword

Sorry for the delay in responding to your comment of 20 Apr 2005
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2005Apr/0038.ht
ml


> I am currently trying to create user-defined spatial operators for
SPARQL.

Interesting!

[...]

> It seems to me that the  current specification does not provide a way
to
pass the description of p1 in the query
[...]
> To support this query, SPARQL needs a mechanism to pass the user
defined
> triples, so they could be added as premises in the query engine (DQL is
> using such concept) prior to the evaluation on any graph.

Note that the FROM keyword is back in the recent draft...
http://www.w3.org/TR/2005/WD-rdf-sparql-query-20050419/#specifyingDataset

which allows specifying premises by reference.

> The handling of
> the premises would be pretty straightforward to implement by any SPARQL
> query engine. 

One implementation strategy that is consistent with the specification
is for the engine to index its dataset before receiving queries and
not handle on-the-fly additional premises.


> ISSUE 2: OPTIONAL OR NULL PARAMETERS.
> ISSUE 3: DESCRIPTION OF SUPPORTED FUNCTIONS IN SPARQL PROTOCOL.

More on those separately...


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E
see you at XTech in Amsterdam 24-27 May?
Received on Saturday, 14 May 2005 04:03:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:14:48 GMT