RDF query IRC meeting of 2003-02-27 - summary

hi all,

Thanks to everyone who came along to the IRC meeting about RDF query
manifest formats - it was very enlightening :)

see

http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-01-15

for raw logs,

http://rdfig.xmlhack.com/2003/02/27/2003-02-27.html

for urls and attendees.

Next meeting:
13th March 2003, 15:00GMT
http://www.timeanddate.com/worldclock/fixedtime.html?day=13&month=3&year=2003&hour=15&min=0&sec=0&p1=0]
on irc.freenode.net, #rdfig

An important conclusion was that we wanted a manifest format that was
flexible enough to be applicable to different syntaxes and different
complexities of query. We are going to start a document about this on
the W3C site. Libby Miller, Andy Seaborne and Alberto Reggiori will be
editors, and the document will summarise discussions on www-rdf-rules
and on irc meetings.

A basic summary of discussions is below. Let me know if I missed
anything off/misunderstood anything.

cheers,

Libby


-----
Summary of the discussion
-----

The discussion was wide-ranging and very interesting, if a bit hard to
follow. We hope that as things settle down a bit we should be able to
stick to the agenda more closely and it'll be calmer.

we did discuss

-----
1.2 What do we intend to do with testcases expressed in this format?
-----

in some depth. Suggestions centred mostly around testing ones own code;
also important was export to other formats, and also comparing QL
expressivity.

There was some consensus on:

* <AndyS> Inference and query can be separate - graph (data+engine) does
inference, then query asks enhanced graph
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-20-31

* <danbri> So, how would folks feel about targetting an RDF query test
format which tried to target the simple conjunctive structure shared by
guha's ql, rdql, squish, algae etc., and which through doing so (i)
helped us find out if there were in fact multiple implemenentations of
this that behave the same (ii) helped us appreciate the value and
implementation difficulty of the various other directions in which rdf
query systems have gone?
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-30-08

(although we were unclear how exactly to do this. There were various
suggestions and we didn't decide on one, or what the difference was
precisely):

 <AndyS> Simple start 1: RDf graph - bnodes, no bArcs
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-40-26
(e.g http://lists.w3.org/Archives/Public/www-archive/2002Apr/0040.html)

<AndyS> Simpel start 2: cwm/N3 LHS
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-40-41-1

<AndyS> Start 3 (not so simple): cwm/N3 rule (formats output)

Jan Grant noted that predicates as variables could mostly not be
required if the RDF database follows subproperty chains.
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-39-00

--
Manifest
--

As far as the manifest itself was concerned, there was a common desire
for something flexible enough to encompass differing syntaxes and
different complexities of query languages:
[[
<danbri> So we note (i) the initial design of the manifest format
targets a to-be-explored common subset of RDFQ, anded triple query; (ii)
we try to keep the manifest format noncommital re concrete syntaxes and
deployment choices...
]]
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-53-28


--
Results
--

We moved to a discussion about results: whether bindings or RDF
graphs were better things to return. Agreement that both could be
useful; and that the subgraph view could be derived fom the bindings:
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-44-05

We discussed resultset formats. There was some agreent on
- RDF resultset format for graph matching
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-56-33
- using bags or collections for results:
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T15-56-19

and then some discussion. I'm not sure we really reached a conclusion:

[[
<AndyS> object of solution is a collection of bindings (one row) as
name/value pairs.
...
<AndyS> yes - binding has variable name ( I don't like that bit but
need to quote the variable so how)
...
<timbl> So a binding only makes sense then in relation to a specific
query. It isn't data which is simply true on its own.
...
<timbl> So the subject would usefully be the query.
...
<AndyS> timbl - subject = query - yes (or query+sources if not in
query)
...
<areggiori> I think we should allow in the manifest rdf:resource for
input source, the query itself and the output - then use
rdf:parseType="Literal" if we want to in-line those in the manifest
]]

--
Several people promised thay would do things:
--

* <areggiori> I will send an email to ww-rules about some alternative
RDF/XML format to Libby's
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T16-00-21

* <libby> I could (a) collect togteher various resulsets formats (b) see
if we decided anything today about manifests
http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T16-07-03-1

* <libby> I also promised to summarise this discussion to the
list http://ilrt.org/discovery/chatlogs/rdfig/2003-02-27.html#T16-07-18

* Libby, Andy Seaborne and Alberto Reggiori to edit a proposal
summarising discusssion on this and future IRC chats and on the mailing
list.

[[
<AndyS> I was thinking not of a page of links but a "working proposal" -
one place for people to comment as working consensus + email for new
stuff
 <libby> the approach we took with the calendar schema, was to annouce
changes to the list, and if anyone creams, rol back. we could do this
with this doc too.
16:22:25  <libby> erk, screams not creams :)
<AndyS> can we divide up the problem space and get small work areas /
task with people to push at them?
]]

* <danbri> AndyS, if you're gonna co-edit this w/ Libby, I'll take an
[action] to get you SSH/CVS access to a shared workspace...
 <libby> well danbri, will you take an action to see if possible for
alberto too, as invited expert maybe?

We also discussed putting tests up on the W3C site for people to use, in
a non-GPLed open source format, probably the W3C's one:
http://www.w3.org/Consortium/Legal/2002/copyright-software-20021231
[[
This formulation of W3C's notice and license became active on December
31 2002. This version removes the copyright ownership notice such that
this license can be used with materials other than those owned by the
W3C, reflects that ERCIM is now a host of the W3C, includes references
to this specific dated version of the license, and removes the ambiguous
grant of "use". Otherwise, this version is the same as the previous
version and is written so as to preserve the Free Software Foundation's
assessment of GPL compatibility and OSI's certification under the Open
Source Definition. Please see our Copyright FAQ for common questions
about using materials from our site, including specific terms and
conditions for packages like libwww, Amaya, and Jigsaw. Other questions
about this notice can be directed to site-policy@w3.org.
]]

Received on Thursday, 27 February 2003 13:12:38 UTC