Meeting 22.10.2013

Participants

Oscar Corcho
Axel Polleres
Stefan Schlobach
Darko Anicic
Jean-Paul Calbimonte
Danh Le Phuoc
Jeff Z. Pan
Marco Balduini
Jacopo Urbani
Alasdair Gray
Lorenz Fischer
Robin Keskisärkkä
Mikko Rinne
Eva Blomqvist

Agenda

Oscar's keynote (as introduction)
RSP C. Group Status
RSP Scope
RSP Use cases and requirements
Future tasks

Pseudo-Minutes

Oscar's Keynote

Find it here [1]

RSP Group Status

Jean-Paul: Where we are now. Short introduction of who are participating.

Jean-Paul: Previous 3 telecons summary: RDF Stream models, query languages and processors: background.

Jean-Paul: CG can create reports and specifications. We aim at drafting specs.

Axel: Industry involvement?

Oscar: e.g. Oracle CEP, how to approach them? Also include triple store providers.

RSP Group Scope

Jean-Paul: Reminder of the Group original goals:

The mission of the RDF Stream Processing Community Group (RSP) is to define a common model for producing, transmitting and continuously querying RDF Streams.
This includes extensions to both RDF and SPARQL for representing streaming data, as well as their semantics.
Moreover this work envisions an ecosystem of streaming and static RDF data sources whose data can be combined through standard models, languages and protocols.

Are we happy with this as scope?

Emanuele: Concerns raised by some Group memebers (e.g. Andy): Processing RDF in streaming fashion? Or processing RDF Streams (notion of ordered/timestamped RDF)?

Emanuele: There might be ambiguity in the name: (RDF Stream) Processing vs. RDF (Stream Processing)

Alasdair: Reminds of the notion of infinite sets of data items, ordered according to ticks or timestamps.

Stefan: Can this be also applied to non-timestamped data?, but such that is so big that cannot be processed in memory and needs streaming data operators?

Alasdair: It is still stored data, the fact that is too large to process does not change the emantics, e.g. an average over the set of values. In stream processing the average computation has different semantics, e.g. compute it over a window.

Oscar: Could be the case, but are we transforming it into a stream then?

Axel: Perhaps this is some sort of "cumulative" stream processing ?

Alasdair: In Stream processing we care about the notion of changes over time.

Emanuele: In principle we considered the continuous processing and not the cumulative one, but can be open if it makes sense.

Axel: CG should be inclusive, these other use cases might have a place in the RSP CG

Mikko: Even for continuous processing, sometimes timestamps are assumed in the RDF Stream metamodel, might no always be the case.

Jean-Paul: Discussion about timestamps astarted in one of the phone calls.

Emanuele: as explained in Oscar's introduction, there are different approaches: one timestamp, intervals, also timestamp per triple, graph, no timestamps.

Oscar: Order might be applied to different dimensions but we are focusing on time for the RSP

Axel, Emanuele: Temporal logics out of scope?

Oscar: yes

Oscar: shows mind map view of scope:

Unordered nature of data
- Small dataset that fits in memory (Ok with sparql) out of scope.
- Big dataset that does not fit, needs cumulative stream-processing of current Sparql: maybe in scope or maybe not
Ordered nature of data (timestamped or not explicitly)
- Recorded stream, it ends: s2r support in Sparql is useful: in scope
- Infinite stream: continuous processing, needs s2r support in sparql.

RSP Use cases and requirements

Jean-Paul: Next step was intended to be working on requirements. Based on use cases we are collecting.

Oscar: Still to gather more use cases.

Axel: Where to collect them.

Emanuele, Jean-Paul: Wiki, template available, can be adapted.

Mikko: First might be important to agree on Concepts, initial draft available.

Future tasks

Alasdair: must get the scope clear

Jean-Paul: Have to write it down and agreee

Stefan: What is the planning?

Jean-Paul: 2014 hoped to have written the specs for model and query language, perhaps too optimistic.

Jean-Paul: Reports and specs forseen: Requirements, RDF Stream Models, Query languages syntax and semantics, REST Services

Oscar: work to be done now: ACTION: still to contribute with use cases.

Oscar: ACTION: Contribute to Mikko's initial proposal of definitions

Alasdair: advisable to have the scope clear by the end of the year.