RDF Data Access Working Group Meeting, April 2004 in Amsterdam

Mr. Hiroyuki Sato of Information Sharing Platform Laboratories NIPPON TELEGRAPH AND TELEPHONE CORPORATION (NTT) attended the meeting as an observer; he has not yet joined the WG.

Minutes

See also: IRC log from Thu, IRC log from Fri.

Convene, review charter and W3C process

Roll call: see attendance above.

Connolly briefly reviewed W3C process, emphasizing that its purpose is to build community consensus to deploy new technologies. Connolly noted that changes to the charter are possible, but they involve negotation with parties outside the WG.

(not presented:formal process model diagram, with versions in.ps,.png, and.svg)

Use Cases

Connolly noted that outreach is an important function of the use cases and requirements document; for example, it would be nice to note a few use cases in a short summary of our work for the WWW2004 conference.

Each member of the WG was invited to speak briefly to their most valued use cases and the connection to their work and their constituency. see IRC log for some details.

ACTION DaveB: write up this educational metadata UC

Clark reported on the current (1.24) use cases draft: some struggles with the "benefits" heading and such.

The group discussed the balance between quality and quantity of use cases: how many do we have time to write up well? Opinions varied from 5 to 10 to 15. Some sentiment favored a small number of elaborated use cases and a larger number of short sketches. External feedback might prompt us to reconsider which use cases to elaborate.

The group discussed wasy to improve the document, including promoting the "problem description" material in the introduction.

ACTION RobS: to generate some RDF (in some form) for 5.1.1.

ACTION EricP: to generate a diagram of a query discussed in the use cases document.

ACTION Kendall: to work with Alberto on 5.1.3.

ACTION Yoshio: to write up "Kate finds an article and wants to find related programs..."

ACTION Kendall: to recast this (5.2.1. Saying things about web resources) in terms of some tool for annotating photos. (DONE in 1.25)

ACTION Kendall: to add negotiation of query lang to candidate requirements. (DONE in 1.25)

The group discussed requirements, especially their relationship to use cases that motivate them.

The discussion got into some design details about styles of query result bindings.

The group discussed requirements for support of various forms of datatyping in some depth. We began collecting straw poll data about support for various requirements.

Connolly constructed straw polls:

if we were to keep only 5 use cases, would this be one of them?
if we were to keep only 10 use cases, would this be one of them? this second question was abandoned part way thru.

The results as reported by Beckett include:

    5  T
1.  8 10 5.1.1.  Finding an email address
2.  9  - 5.6.1.  Finding information about motorcycle parts
3.  8  - 5.5.1.  Finding unknown titles (media)
4.  7 12 5.1.3.  Monitoring news events
5.  7  - 5.8.1.  Avoiding traffic jams
    7  - 5.2.2.  Discovering what people say about news stories
   
    6  - 5.4.1.  Exploring my neighborhood

    4  7 5.2.1.  Saying things about web resources
    4  - 5.7.1.  Finding input and output documents for test cases

    2  7 5.9.1.  Ordering an x-ray
    2  - 5.8.2.  Finding the cheapest flight from Boston to Chicago
    2  - 5.3.1.  Tracking accounts and customers
    2  - 5.10.1. Find employees by group
    1  4 5.1.2.  Regularly executing a query
    1  - 5.7.2.  Describing software configurations

The straw polls do not take into account the following use cases, which had not been digested into the use cases document or, in some cases, not been written:

Browsing Example - tell me about X, ACTION Steve Harris
Bandwidth Efficiency - bytesize of resultset, ACTION SteveH
Chunking" - limit and offset, Steve Harris
OWL Ontology Browser, Pat Hayes

ACTION Kendall: to look at Yoshio's use cases.

ACTION Janne: write a use case to motivate the requirement of having queries written in RDF. (WITHDRAWN?)

The group only briefly discussed a plan for publishing the use cases and requirements document as a W3C first public working draft. The 3 month "heartbeat" requirement means we need to publish in May; May 17, the start of the WWW2004 web conference, is an interesting, if challenging, target.

On Friday morning, the editor reported some progress on updating the document (Revision 1.25 2004/04/23 08:13:40). The WG RESOLVED: kendall has done a great job.

ACTION: AndyS, EricP to review kendallC draft by 4 May telcon

As discussion of datatype constraints brought up some implementation experience with specific cases, the chair asked for a volunteer to start collecting test cases.

ACTION SteveH: to draft and maintain a list of test sketch cases, including x < y, x < 18.

As recorded in requirements straw polls, various members of the WG supported or opposed various requirements:

Queries with optional triples (straw poll on required: yes 9, no 1)

Queries with disjunction (straw poll on required: yes 4, no 3, chat more 2-3) UC: 5.3.1
Queries expressing arbitrary RDF data types (straw poll on required: yes 6, no: 3)
Queries supporting datatypes xsd:integer and xsd:float with operations: less than, equal, greater than [DATATYPES-IF]
Queries supporting date and time datatypes with operations equal, before and after. [DATATYPE-DT] (straw poll, yes: 9, no: 3)
Queries expressible as URLs. (straw poll: yes 9, no: 1)
Queries serializable as text. (straw poll: yes everyone)
Query results in user-selectable result template. (straw poll: yes: 0, no: 10)
Query results in user-selectable Internet Media Types. (straw poll: yes: 5, no: 4)
Query results as a sub-graph of the queried RDF graph (straw poll: yes 10, no: 2)
Query results as multiple solution graphs (straw poll: yes 4, no 5)
Queries for the non-existence of one or more triples in a graph. (straw poll: yes 6, no 1)
Queries expressible in a syntax that is easily read and written by people. (straw poll: yes 10, no 1)
Queries shall allow execution against a local RDF storage service without network support (i.e., queries independent of any network protocol). (straw poll: yes 10, no 0)
The protocol shall allow the negotiation of query languages (straw poll: yes 4, no 8)
Query results can be returned as 0 or more variable bindings (tuples) (straw poll: yes 12, no 1)
Query results must return source/provenance (straw poll: yes 2, no 10)
Query results may return source/provenance (straw poll: yes 8, no 3)
Query results can be interpreted without knowledge of RDF. (straw poll: yes 2, no 6)
Queries written in an RDF syntax (straw poll: 0, no: 8)
The graph of the query looks like the graph of the result (straw poll: yes 1, no 8)
The query language shall be extendable (such as via function calls or namespace mechanism) (straw poll: yes 11, no 1)
The query protocol shall address bandwidth issues of result formats. (straw poll: yes 9, no 1)
The query language/protocol supports limit, offset and ordering of results. (straw poll: yes 5, no 3)
The query language/protocol supports limiting of the number of results. (straw poll: yes 9, no 2)

Obvious:

Queries with paths of two or more edges

Insufficient Support:

Query results using 1) graph entailment *OR*

Query results (not using entailment) but 2) treating the graph as a fixed object

Query results can be returned as triple bindings

The query protocol should be cognizant of network access.

Web Architecture in Brief

Connolly presented some slides on the Web Architecture document and the TAG. Blah blah blah.

Meeting Schedule

As recorded in DAWG telcon and F2F meeting decisions, after discussing various constraints, we made the following decisions:

Resolved to have telcons: 14:30Z Tuesday (10:30 Boston / 11:30pm tokyo / 16:30 Paris / 07:30am SF) starting 4th May 2004
Abstained: RS
Resolved to meet 2004-07-14/15 Network Inference place: California
tentative regrets: JS, DB
Resolved to meet 2004-09-16/17 HP AndyS place: bristol
RS abstaining
Resolved to meet 2004-12-01/02 Network Inference place: Las Vegas
JdR abstaining

Survey of existing technologies

The first session consistend of three 10-minute talks followed by an hour of discussion.

on the use of XPointer and the HT TP "Range " header as a possible scalable and extensible RDF data access protocol by Bryan Thompson (remote)
XQuery: a whirlwind tour by HowardK
Using XQuery for RDF: costs and benefits of one approach by RobS
@@materials

see the IRC log from Fri for some notes on discussion.

ACTION DanC: inform the TAG of the conflict between sec13.9 of the HTTP spec and TAG's recommendation on issue 7

In the second session, Kendall's Versa talk was preempted by more work on the usecases document, and Pat Hayes's DAML QL talk was posponed until Pat is available, so we had 2 short talks on implementation experience with RDQL:

Implementing RDQL in Rasqal by DaveB
RDQL: an implementation experience by Alberto

After a short discussion, the WG resolved to thank the hosts, with applause, and the meeting was adjourned.

RDF Data Access Working Group Meeting, April 2004

hosted near Amsterdam by @semantics

Venue

Participants

Agenda

Thu, 22 Apr

Fri, 23 Apr

Presentation Materials and Meeting Records