W3C home > Mailing lists > Public > www-rdf-rules@w3.org > November 2003

Re: RDF Query -- possible WG charter draft for discussion

From: Alberto Reggiori <alberto@asemantics.com>
Date: Mon, 17 Nov 2003 18:53:50 +0100
Message-Id: <FFED31F9-1926-11D8-87DD-0003939CA324@asemantics.com>
Cc: 'www-rdf-rules@w3.org' <www-rdf-rules@w3.org>, Eric Prud'hommeaux <eric@w3.org>, staff@asemantics.com
To: Dan Brickley <danbri@w3.org>

Hi Dan,

great to see this happening!

We certainly would support the open and general RDF Query WG
charter/initiative call [1].

And we think that getting the RDF query right is key to the success of  
the
whole Semantic Web; we also strongly believe that such work should focus
on getting there soon (using mostly what we already got up and running
today in existing tools) and let a second layer of the standard/work  
grow
gracefully later i.e. "quick and dirty" vs. "complete and never
implemented" or "80% of problem with 20% solution" :) We believe that  
at this point
the comminity has enough implementation experience to get this first  
layer
right in a short period of time.

As most of our concerns at the moment are about RDF query
interoperability between existing Web software solutions/tools and its
standard database interfaces such as JDBC, ODBC and DBI, we feel like
as important for any upcoming RDF Query standard to use and reuse as
much as possible existing work done in the field of RDBMS/SQL and in
particular from XML based systems using XPath, XQuery and XSLT. We
found out this as being a big success factor for convincing people,
customers and end-users to use/buy the RDF machinery instead of scare
them away ;) So, SQL <---> RDF <---> XML interoperability is relevant
here.

Of course RDF will need more expressive query languages than simple
XPath/XQuery or SQL ones. And as already outlined into the charter
document, it is also important that any sensible RDF Query language
(model and syntax) must have at the core some extension mechanism which
will let applications and code grow gracefully in domain specific areas
if needed. With this we mean that the core model and spec should allow
a very simple and effective extension mechanism which would allow third
parties to build on top of a core RDF query machinery more advanced
facilities e.g. filters, adaptors to legacy databases or Web services.
We believe that any of such extensions will be needed as soon as some
realistic RDF application is built, especially if any realistic and on
real-life dataset is used.

Having said that here is a rough list of requirements we would like to
see considered in the RDF Query work:

1. the core query language machinery must be as simple as possible to
implement and based on the current RDF (recommended) model

2. it must support a simple pattern-matching feature to "grep" the
input RDF graph sensibly - a part simple conjunctive triple-patterns
this might also involve some free-text filters or simple constraints

3. it must accommodate a simple and modular extension mechanism to
plug-in any application or architectural specific facility e.g. by
using specific XML-Namespaces bound to specific extensions or some
event-driven interface ala SAX (or any other extension mechanism) i.e.
"pipe" the matching results to the next phase/step or something

4. it must support (canonical) RDF/XML, XML and SQL TABLE like result
sets (whole sub-graph or graphlets) - or any sensible transformation of
those (negotiable)

5. the query language syntax should be simple and as much familiar as
possible to end-users and developers and easy to pick up e.g. SQL or
XPath (we are in favor of an SQL-like augmented syntax ala
RDQL/SquishQL (+ XPath-ish eventually) which can easily fit into
existing DB infrastructures and can exploit a lot of 'extra's in the
SQL world such transactions, views and so on - this also kind of
require the result set to be a "traditional" SQL table (rows/columns)
but it might not be the case in the RDF world)

6.  conversion between different language syntaxes should be possible
and trivial - and ignore extensions not implemented or not
understandable to the application possible too without breaking the
whole system i.e. "grey out some rules"

7. it should support/have some basic core RESTful discovery and remote
access query protocol [2] i.e. access mechanism to RDF data and
metadata - where the source RDF might reside on different Web sites,
different organizations and having different responsibilities and so on

8. have built in support for XML-Datatypes i.e. understands about basic
scalar data types (e.g. int, float, double, date) and expect
applications to do so

Possible extensions which we might be interested to see implemented [3]:

1. support for RDF provenance/contexts (but this could also fit in the
RDF Query core due is widely used and necessary in applications today)

2. support for optional matches or might bind variables/branches (which
might as well fit into the core language if necessary)

3. in case the result set is a real ER-like TABLE:
	--> ordering of results
	--> recursive queries like nested SELECT statements

4. simple GIS and geo-located services modules:
	--> retrieve results/triples based on their space/temporal location
	--> search by geographic coverage, for example, to select information
about cities in a given latitude/longitude region.

5. support some kind of UPDATE/DELETE feature to easy RDF data and
metadata management

Concerning the query requirement phase, we feel like as useful starting
points the work already done in the past [4] [5] - which lists clearly
most of the requirements for such RDF Query language proposal and will
help to consolidate and document existing languages and systems. We
invite people on this list to submit/add more use-cases and scenarios,
especially if these are coming from real-world applications or customer
requirements :) Feel also free to post comments to this list about
existing syntaxes, use-cases and example queries - this will also help
to better understand where to go next or not.

The set up and maintenance of a interoperability test repository for
RDF Query (and Rules) RDF files and queries such as [6] will also be of
relevance for the RDF developer community. As soon as more RDF tools
will come on the market, a set of reference test cases and test dataset
sets will be of vital importance for testing and tuning. Along those
lines some work was started to fill up a Web site (and CVS repository)
on SourceForge [7] - we feel like relevant such an initiative and would
like to support that in the upcoming months, especially before
the real RDF Query WG would start.

Yours

Alberto

[1] http://www.w3.org/2003/10/RDF-Query-Charter
[2] http://www.w3.org/Submission/2003/SUBM-rdf-netapi-20031002/
[3] http://lists.w3.org/Archives/Public/www-rdf-rules/2003Apr/0031.html
[4]  
http://rdfstore.sourceforge.net/2002/06/24/rdf-query/query-use- 
cases.html
[5] http://www.w3.org/2001/11/13-RDF-Query-Rules/
[6] http://www.w3.org/2003/03/rdfqr-tests/
[7] http://sourceforge.net/projects/rdftests

On Nov 7, 2003, at 7:28 PM, Dan Brickley wrote:

>
> Hi folks,
>
> In the years since W3C's QL'98 workshop on Query languages
> for the Web [1], there have been an increasing number of RDF
> query languages and implementations (hence this RDF IG mailing list).
> Many RDF-based applications use one or more of these languages
> to express queries against repositories of RDF data.
>
> The W3C Team, with input from the Semantic Web Coordination Group,
> is currently drafting proposals for a 2nd phase of our
> Semantic Web Activity. One area of possible work we are considering
> is RDF Query. This list has been pretty active in discussing
> the details of RDF query, its relationship with RDF rules, XQuery etc.,
> as well as hosting practical interop collaboration, so we're
> circulating an early discussion draft for comments.
>
> Did I mention yet that it is an *early* *draft*? For **discussion**?
> Nothing is set in stone. Specifically, we haven't proposed anything
> yet to W3C's Advisory Committee, and they've not approved anything.
>
> The current draft is at [2]. EricP did most of the work on it. It will
> quite possibly change over the coming weeks, so please be sure to cite
> the $Revision number from the 'Status' section. The current version
> as I write this is 1.19.
>
> Issues we are still discussing include the schedule, relationship to
> XQuery, OWL etc., and various other things flagged up in the document
> (CSS styled in Red at least in my browser). We are also interested
> in feedback regarding the extent to which mechanisms for remote access
> (ie. Web service interfaces; protocols) should be addressed
> by any standards-track Working Group we charter.
>
> We would appreciate your comments by 20 November 2003, preferably to
> this (www-rdf-rules@w3.org) public, archived([3]) list. A separate
> discussion draft for work in the Rules area is also in preparation. The
> relationship between rules and query is an area we particularly
> welcome implementor feedback on. The thread earlier this week([4])
> was both interesting and helpful btw.
>
> Thanks for your input!
>
> cheers,
>
> Dan
>
> [1] http://www.w3.org/TandS/QL/QL98/
> [2] http://www.w3.org/2003/10/RDF-Query-Charter
> [3] http://lists.w3.org/Archives/Public/www-rdf-rules/
> [4] http://lists.w3.org/Archives/Public/www-rdf-rules/2003Nov/0000.html
> --
> danbri@w3.org RDF Interest Group chair
> http://www.w3.org/people/danbri/
Received on Monday, 17 November 2003 12:54:01 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:53:10 GMT