RE: Innovation, community and queries

> There are several good proposals for QLs out there, and I think 
> that creating a "working draft QL" hybrid of a couple of these 
> would be a seriously good idea. 

This is a good way forward - take the first generation languages and
consolidate the ideas.

It would really help toolkit/framework builders, both current ones and
future ones to have one common QL.  It would help semantic web application
builders to have a common language they could get to learn.  It would help
tool builders to be independent of the RDF system they would otherwise need
to choose.

	Andy 

-----Original Message-----
From: Jeen Broekstra [mailto:jeen.broekstra@aidministrator.nl] 
Sent: 23 May 2002 13:15
To: Graham Klyne; Aaron Swartz
Cc: www-rdf-interest@w3.org
Subject: Re: Innovation, community and queries



Graham,

On Thursday 23 May 2002 01:40, Graham Klyne wrote:

> As someone who has recently designed and implemented a (yet
> another) RDF query "language", I'm not convinced we're ready to 
> standardize. I'm not convinced we know enough about the performance 
> issues in RDF, and I'm also not convinced that standardizing a query 
> language at this time would bring great benefits.  But I could be 
> wrong on both counts.

A possible benefit I see of standardizing on an RDF QL would be 
to harness efforts made in several groups to optimize tools for 
"their own" QL into a single framework. This makes it much 
easier for such groups to compare results and benefit from each 
other.

There are several good proposals for QLs out there, and I think 
that creating a "working draft QL" hybrid of a couple of these 
would be a seriously good idea. In fact, we have some plans of 
our own to try this, but we haven't pursued this further sofar, 
because we do not think it is a good idea to propose Yet Another 
RDF QL[tm] on our own.

As for performance issues, I do not think that should be an 
impediment to standardizing on a language, or at least making a 
"starting effort" to do so. The implementation and storage 
format should be seen as seperate from the QL spec, I think. I 
would say that expressive power and complexity/"learning curve" 
are the most important parameters for any QL.

> My own intuition is that a query language for RDF should aim to 
> operate at a higher level than "find this pattern of triples", but in 
> my implementation it was hard to break away from.

I am not quite sure what kind of higher level you have in mind. 
Do you mean something like RDF Schema semantics interpretation, 
or something more along the lines of query formulation in 
natural language?

> I'd like to see more work on storage formats before we
> nail down a query language.

As I said, I think they are seperate issues. I'm hammering this 
because the system I work on (Sesame) operates on the premise 
that storage and retrieval are completely abstract operations. 
The query interpreter does not have knowledge of the storage 
format, nor is the storage format dependent on the QL we use.

Regards,

Jeen
-- 
jeen.broekstra@aidministrator.nl
aidministrator nederland bv - http://www.aidministrator.nl/ 
julianaplein 14b, 3817 cs amersfoort, the netherlands
tel. +31-(0)33-4659987, fax. +31-(0)33-4659987

Received on Thursday, 23 May 2002 08:42:30 UTC