Notes from RDF Query BOF, SWWS, 2001-08-01

Libby Miller

see also text version and RDF metadata and the TMQL list

[to start with we wrote up sample queries on the whiteboard for 4 of the 8 query languages in the room (see photos from the RDF Query BOF)]

Eric Prud'hommeaux led the classification of the query languages into:

• subgraph matching
• "reachable" (keep following arcs out of a node - David Allsopp's implementation does this)
• + subgraph constraints on reachability
• + a special case of this is transitive closure (e.g. follow subclass arcs only)

EP: Reachable is like a transitive query with any predicate more specialised is transitive query with any predicate, like transitive closure with subclass and subproperty.

didn't catch his name: what about rules?

EP: none of these have assertion capability. where do assertions go?

• the original dataset? (changing the set)
• a working set?

didn't catch his name: might want to select from a set of rules as well as from a set of RDF documents

[then we went around the room, each of us saying our names and why we were interested in RDF query.]

Libby Miller: Squish, Inkling: simple, for testing, for scripters

JB:
Sesame, one of two implementations of RQL.
An RDF schema language
quite complex but powerful - native support for RDF schema semantics
not querying triples but the graph.
path expression syntax
Ontoknowledge project partner - a requirement to be able to query the ontology.

EP:

0 --> O --> 0

vs
0 --> 0
and
0 --> 0

?

JB: yes

Michel Klein:
want a query language making the semantics useful
for scripting, an instant QL

EP: scenarios in mind?

MK: PHP

Fernando Arguello:

An undergraduate at Stanford. Wanted to learn about RDF query and
specifically efficient RDF query language implementations

Don Johnson:

Helps people map out requirements for requirement engineering
His organisation used SQL for years, but the database was too controlled and prevented people from expressing their ideas fully.
Now they are using a semantic web solution. This is interesting, but it takes a lot of time to go through the data. Aim is therefore to speed requirements analysis

They have a simple query language which is inefficient but better than nothing

didn't catch his name:

a grad student at Stanford
interested in efficient query

H Holger Rath:

works for Empolis
coeditor of TMQL
collecting requirements for TMQL at the moment, hopefully approval soon.
waiting for the data model group
interested in discussing requirements

EP: figuring out the relationship between RDF and TM is useful

Chutiporn Anutariya:

Asian Institute of technology - PhD student

working on a framework for modeling and manipulating RDF data. Has a
query language with an XML syntax.

David Jones:

Boeing
part of the 'user community' for RDF query.
has lots of data not in RDF, which needs to stay in that form.
wants an RDF query language which will do query transformations getting
instance data from relational databases using schemas
also would like to be able to query any XML document
use RDF to get answers over different RDF data sources.

Sarah Comai

works on databases and XML query languages.

Pierre-Antoine Champin

working on case based reasoning where the case is an RDF graph.
cases where subgraph matches the pattern
interested in matching similar graphs rather than exactly matching graphs.
an assistant system.

didn't catch his name:

student at Maryland working on the DAML project
has a query language - rules based
mostly interested in storage issues

Matthias Palmer

Royal institute of technology Stockholm (KTH)
human-computer interactions
mind maps, concept maps -> RDF
transitivity, inheritance of metadata, graphical interface

Todd Siegel

• likes good questions, fond of language
• e.g. aggregating people by similarities in their tastes
• 'tasty matrix'

David Allsopp

works for QinetiQ, which was the Defence Evaluation and Research agency
in the UK legacy systems -> RDF
has a QL, in memory triple store
wants a realistic way of storing the data if untrustworthy
relates to query because need to be able to use this to pick out the set
to query.
stored data reified but doesn't want to use quads to query it
message-based
you can get complete docs for context
efficiency

Eric Prud'hommeaux:

W3C
writing RDF, semantic web tools 2 years
Algae is the 'hammer I run round with and see everything as a nail' :)
e.g. Annotea
Algae is an API and a data structure and an s_expression query language
can do assertions
general triple store, in-memory triple store, general SQL access

[end of intros]

DJ: what is the status of RDF query?

EP:
here's the W3C process:

• membership say we want stuff
• a F2F meetign is held for interested parties
• call for participation goes out to the members
• members vote yes or no
• -> activity
• strawman proposal, constraints, requirements
• WG - evaluate existing ones, crate new one maybe
• results go to membership and then director

none of these things had been done get for RDF query
often want to go through the process if divergence will hurt the industry

DJ: what good is RDF is you can't query it?
query is very useful

DA: API stuff, but yeah

EP: sources to query?

• http resources
• local files
• triples stores, could be relational databases
• general RDB - not a triple store
• arbitrary xml
• API, messaging

DJ: trust needs to be embedded deeply
there's a difference between SQL which is tightly controlled

EP: yeah, what I call a 'homogenous trust environment'

- same level of trust
trust is part of the selection criteria. It comes from the sources - is defined in the sources

DA: tagging internal data yourself

EP: Algae can get quads out - you can say, I only want to get things from one source

DA: 2 different levels - query external databases and query own databases

EP: reification and quads
quads or reified - same
4th is often called the attribution or context
dereferencing the statement ID can get the attribution in Algae -
especially useful for Annotations - http credence

HHR:

[gives brief introduction to topic maps]:
topic as central point
address it directly via url of subject
if not addressible have to invent an addressible thing to represent it
Xlink - potentially 2+ links to the same subject

• updates?
• contents?
• requerying results?

EP: I'll find/set up a list for RDF query

Participants