W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > October to December 2004

Minutes of DAWG telecon 2004-10-26 for review

From: Tom Adams <tom@tucanatech.com>
Date: Mon, 1 Nov 2004 10:23:01 -0500
Message-Id: <EB6BA66B-2C19-11D9-9C17-000A95C9112A@tucanatech.com>
To: DAWG list <public-rdf-dawg@w3.org>

Minutes of RDF DAWG teleconference 2004-10-26T14:30Z for review


IRC log:

Scribe: Tom Adams


1. Convene, take roll, review record and agenda

   Steve Harris
   Kevin Wilkinson
   Eric Prud'hommeaux
   Dan Connolly
   Pat Hayes
   Tom Adams
   Farrukh Najmi
   Simon Raboczi
   Alberto Reggiori
   Dave Beckett
   Jos De Roo

   Andy Seaborne
   Hiroyuki Sato

RESOLVED: to accept telcon minutes
as a true record.

Next meeting 2004-11-02 14:30 UTC
Regrets: ?
Chair: Dan Connolly
Scribe: Alberto Reggiori

All following actions continued:

ACTION KendallC: expose our walking tour data to SPARQL querying clients
  ETA before F2F4
ACTION EricP: find logistics re F2F5 at tech plenary in Boston March
  28 Feb - 4 Mar (2 days).  Some WG preference to Mon/Tue of that week.
  some progress:  
ACTION AlbertoR: owns issue 'DESCRIBE'
ACTION SteveH: own (i.e. propose resolution to) disjunction issue
ACTION SteveH: owns issue 'nested optionals'
ACTION DanC: owner of issue 'yes or no questions'
ACTION DaveB: Update the source section 9, add more formal links,
update the examples, try to think about extra constraints as EricP
proposed (SOURCE ?s and ?s only in in SELECT). Look at various
people's source test cases.

New action summary:

ACTION: Eric to ask Kendall to put use case into UC&R or find a use  
case that covers it
ACTION: SteveH to take rs:size out of expected results from all tests
ACTION: EricP to supply definitions for SELECT (vars ordered or not?)  

2. Web Services Constraints and Capabilities

ACTION EricP: draft UC on overlap between RDF query and web service
constraints with respect to WS-Policy

Eric wrote something up [1], cited example, unsure of expressivity.


Example query for web services constraints, written by EricP (DaveB:  
page is unsigned). The matrix on the page is familiar to WS policy  


Expressing what you need to use our web server, must use a security  
token, X509, etc. Expressed as simple value disjunction for this use  
case. Not all use cases being kicked around in WS meeting can be  
covered by value disjunction.

DaveB: replace with ... { ?assertion wss:tokentype ?t} AND (t =  
"wsse:Kerberosv5TGT" or ?t = "wsse:X509v3" )
DavdB: ^- value disjunction

AlbertoR: simple disjunction as motivated by  
0000.html use case

Another use case was offered: You need X509 or (SSL and basic auth)

DaveB asked a question of Eric last week on the decisions of the  
ws-policy ppl.

A quick straw poll was conducted on whether to stick this into use  
cases. 6 came back in the affirmative.

DaveB: If they stopped at disjunction, why did they?

DaveB: I heard eric say they were about 50/50 for adding/not adding  

Another quick straw poll was conducted on not bothering to include in  
the use cases, 2 people in affirmative.

Dave and Steve suggest that we have enough use cases.

ACTION: Eric to ask Kendall to put use case into UC&R or find a use  
case that covers it

ref to steveH's proposal on disjunction  

SteveH: ref to my comment on CNF:  

3. Feedback on the SPARQL design

Dirk provided details about DB interfaces, re '?' and '$'

Note. Eric acted as editor on Andy's behalf

A quick strawpoll was conducted into the variable prefix character,  
summary below.

SteveH: No pref
KevinW:  no pref
SimonR: no particular required prefix character, $, ?, :. Specifically,  
I liked the suggestion where there was no required prefix character.  
And identifier that wasn't a qname (i.e. no colon) would be a variable.
DanC: No pref
patH: $ (patH: Suggest a design in which the user declares the  
character, but has a default, which i $)
TomAdams: $
FarrukhNajmi: $
AlbertoR, $
DaveB, $
JosD: ?

ericP Sensed the way the wind is blowing and changed to $

SimonR: I agree with Farrukh about having either one or the other, at  
AlbertoR: can live with both - but preference (best practice) is $
SimonR: I sort of prefer $ because it's more like XPath/XQuery/XSLT.

Cont. action: PatH was to reviewed SPARQL def'ns post-publication

Dan Connolly (Tuesday, 19 October)
definitions for SELECT, projection, substitution [was: [Fwd: Re: ...]]

DanC: Will leave as input to editor, update as they see fit.

patH: Thought it was on hold, will do it this week.

DanC: Pat to work from editors draft at  

DanC sent defns for select (above), editor didn't read them...

4. PREFIX syntax
cf http://www.w3.org/2001/sw/DataAccess/issues#prefixSyntax

ACTION: DaveB to illustrate prefix interaction details by example/test
and following thread, including a proposal:

DanC: Thinks we should make the language smaller.

DaveB: example  
DaveB: proposal  

AlbertoR: Didn't digest grammar, like proposal, is this related to  
serialisation of results

DaveB: Not aware of relation to result

AlbertoR: wondered if the proposal/solution would cover things like  
select ?prefix:var ?prefix:var1

ericP : My position on the three-part proposal: +1 -2 +3

DanC: (hmm... why did I make prefix syntax a WG issue and not other  
syntax details? chairing error.)

Eric P gave his decision, accpt 1 and 3, but not 2

Simon is for just doing prefixing at the beginning

SimonR: Now that I've read 164, I'd support part 1 and 3, but not 2. 2  
is a special case which changes the consistent rule elsewhere that  
qnames and URIs are interchangeable.
SimonR: +1 to EricP's position, as I understand it.

Dan opted to leave this as a he who does the work writes the rule, is  
suggesting a use case

5. Toward adopting some tests

DaveB: If ? is chaging, all the tests are unapprovable

AlbertoR posted about some SOURCE tests details - i.e. my experience  
while running tests

AlbertoR: me too - I have the machinery in place - it does not do  
OPTIONAL though - results are checked "manually" to - but could use  
some propose isomorphism utility

DaveB: DanC walsk through item 5, looking at dawg-triple-pattern-001

DaveB: has run it. steveh has translated to rdql and run it. josH has  
done similar ot n3

AlbertoR: and Alberto is about to run some simple tests (a part the  
SOURCE ones) - but OPTIONAL ones

DaveB: Not redundant if you have optional triples

DanC will think about results for optional triples

?? Not happy with result format, does p come before q?

AlbertoR: does SQL / ER models "demand" order on columns? not sure...

DaveB: "query results are a set of pattern solutions." -- sec 2, SPARQL  
paragraph 2

SimonR: There is no order specified, so the ordering does not matter in  
this case.

?? Sometimes order matters, such as in XSLT

SimonR: Points to  

DanC: Suggests that the protocol does this, not SPARQL

AlbertoR: right - but XSLT is already outside the SPARQL protocol stack  
I expect...good to sort RDF-for-XML syntax for example (we do that  

A pattern solution is a set of bindings (no order), the results is a  
set of query solns

The order of a given row is not order, the order of the rows is not  

DanC: points to section 8:  

Spec says:
"Results can be thought of as a table, with one row per query solution.  
Some cells may be empty because a variable is not bound in that  
particular solution."

AlbertoR: yes - I can read (and implemented) un-order result sets (rows  
or columns) - just a set of results

SimonR: Have redefined "results"

Simon is encouraged to look at wording and defn

AlbertoR: References earlier comment.
AlbertoR: Need to spell out clearly that we don't support ordering.

DanC: Expects result format to be in the protocol level, would expect  
results to be deterministic.

DanC makes comments that for protocol format, columns are ordered  
lexigraphically. Rows are a separate issue.

DanC: Asks Steve about taking size out of result format

ACTION SteveH: take rs:size out of expected results from all tests

ACTION EricP: supply definitions for SELECT (vars ordered or not?) ala  

ericP: Points on barrier between protocol and query language.
ericP: QL specifies select, gives them back in order specified, rows  
separate issue.
ericP: In terms of getting information fro QL spec to protocol spec,  
can the protocol spec peak into the QL?

DanC: See similar issues, wants to see what Eric comes up with

ericP: Talking about removing the number of rows. If you remove the  
number of rows and do a graph compare between an implementation and  
that defined in the spec, if you don't have a distinct, won't be able  
to automatically merge the tuples that have the same binding.
ericP: May be able to do it with a rule, but not tailored to the result  

DanC: The merged example graphs (texas) are isomorphic

??: Graph are logically equivalent

ericP: foo rs:result [ rs:variable "where"; rs:value "Texas"];
ericP: rs:result [ rs:variable "where"; rs:value "Boston"];
ericP: rs:result [ rs:variable "where"; rs:value "Texas"].
ericP: compare to:
ericP: foo rs:result [ rs:variable "where"; rs:value "Boston"];
ericP: rs:result [ rs:variable "where"; rs:value "Texas"].

DaveB: patH says they are not isomorphic, but are logically equivalent

DaveB: (rdf-entailed I guess)

The discussion continued on this topid, the meeting was adjourned at  


Tom Adams                  | Tucana Technologies, Inc.
Support Engineer           |   Office: +1 703 871 5312
tom@tucanatech.com         |     Cell: +1 571 594 0847
http://www.tucanatech.com  |      Fax: +1 877 290 6687
Received on Monday, 1 November 2004 15:23:04 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:00:45 UTC