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

my notes from today's 2nd session of BOF on RDF Query at WWW-2003, largely discussing XQuery relationship

From: Benjamin Grosof <bgrosof@MIT.EDU>
Date: Fri, 23 May 2003 14:15:02 -0400
Message-Id: <5.1.0.14.2.20030523141030.01d7c4f0@po12.mit.edu>
To: www-rdf-rules@w3.org
Cc: eric@w3.org, connolly@w3.org, massimo@w3.org

Hi folks, here are my notes from today; it was a bit of a struggle.  Please 
post corrections/additions to the list.
-- Benjamin

% 2nd session of BOF on RDF Query:  Fri. eve 5/23/2003
% notes (non-official!) by Benjamin Grosof

proposal by Jonathan Robie:  Syntactic Web

Liam Quinn, XQuery person, is attending

DanC:  can write an RDF parser in XQuery

TimBL: [part of an] answer to Paul Cotton's request about
what RDF req's to XQuery [could be]:
we need RDF parsing functionality to be efficient

Liam:  not well understood:  XQuery doesn't search XML,
rather it transforms (ordered labelled directed) graphs into graphs;
the input graphs can be forests and can have cycles using IDREF's;
doesn't use XML syntax either

uses XPath model, then extends it by adding
about 10 keywords and some other stuff

XQuery can set an unordered flag in effect
-- i.e., there's a function that does a non-deterministic permute,
that XQuery engines notice and realize from an optimization viewpoint
means that order is to be ignored

thus essentially XQuery can handle as input and output
the full expressive class of graphs of RDF triples

can use a holds(subj,pred,subj) style
to represent RDF triples, write queries that way

Benj (with Liam nodding yes):
can compose/chain queries/views or inferencing engine results
e.g., to do RDFS subclassof, domain, range, type stuff;
but probably more complex to do reification and classes as instances

DanBri:  there are some other ways people do this, e.g.,
holds_OWL

i.e., we can accomplish the fuller RDFS semantics of entailment
via inferencing schemas represented in XQuery that way

thus we can factor this issue, and get back to considering
just queries at level of RDF triples

EricP:
1. implementation shown by Liam Quinn -- with //path and //holds
(and using Galax (spelling?))
only worked on a custom serialization of RDF;
2. such a function to parse real RDF-XML would probably be difficult/clumsy

Benj Q: why (2.)?
(queued)

Liam:  Jonathan Robie wrote something that does (2.)

EricP:  another issue to discuss is the suggestion to replace XPath
with RDFPath [cf. Andy Seaborne remark in first session of BOF]

Alberto Reggiori:  have you tried this with JDBC/ODBC to a lot of sources?
since it seems important to try to do this
A by Liam:  no

Liam:  there are components that convert XQuery into SQL, however

Alberto:  I'll be happy if I can do XSLT back and forth

Dirk-Willem Van Gulk:  JDBC/ODBC actually can do more than SQL,
since on the wire is UTF-8 not SQL per se;
thus can actually push some of RDF-Query
or XQuery kind of stuff into/through them

EricP:  why use XQuery, again?

DanC and DanBri:  because they got there first

Benj Q:  do we have technical consensus that RDF triples can be encoded

Liam:

it's really the XQuery data model not XML encoding

the data model includes some basic operations

basic operations:  getnode, getchildren, getnodevalue, etc.

then above that level are functions/operations:
set-ish (union, intersection), usual arithmetic operators

DanBri:  it's not really exactly equivalence we're talking about
between the RDF graphs and the XQuery data models, but ability
to reduce one to another

Benj:  yes, it's about reducibility

TimBL:
yes, but not reducibility of the RDF *Query language*;
1. it appears that we can map RDF data model into XML data model,
and vice versa
2. and use XQuery machinery as a way to implement RDF Query yet to be
(that we have a pretty good idea of so far)
3. but XPath syntax is pretty clumsy;
4. despite that "to sweeten the pill" for those familiar with XQuery,
we might make the RDF Query lang's syntax look
relatively similar to XQuery's, tho' it would mean something somewhat
different,
5. and would get transformed into something considerably more complicated
if transformed into XQuery

candidates for where to find functional components to build
RDF query-answering engines:

Timbl:  actually closer mapping to SQL, since
they're is relational and unordered, than to
XQuery engines

Benj:  what about implementing path expressions, aren't XQuery engines
better than SQL engines for that?

EricP and Steve:  with all the indexing the SQL engines do, quite possibly
/ probably  not



________________________________________________________________________________________________
Prof. Benjamin Grosof
Web Technologies for E-Commerce, Business Policies, E-Contracting, Rules, 
XML, Agents, Semantic Web Services
MIT Sloan School of Management, Information Technology group
http://ebusiness.mit.edu/bgrosof or http://www.mit.edu/~bgrosof
Received on Friday, 23 May 2003 14:13:39 GMT

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