Re: What is a stream element? (was RSP next calls)

p.s.: It seems to me that this aspect of varying size objects (those with the same timestamp) within a stream brings together the two different apsects of streaming we seemd to not get together in the meeting at ISWC… let me try to illustrate it with a picture:



|   ordered RDF triples  --->   t1[2013-11-22T12:59], t2[2013-11-22T12:59], t3[2013-11-22T12:59], t4[2013-11-22T12:59]               
|   ordered RDF triples  --->   t1[2013-11-22T13:00], t2[2013-11-22T13:00], t3[2013-11-22T13:00] ...
|   ...
|
V 

What I want to illustrate here is the difference between streaming *temporal* RDF data and *stream processing* of RDF data *with the same time stamp*, but that these may actually occur side by side and not be treated in isolation.

In fact, one might define "event" here as triples connected in such a two dimensional stream via different time points, whereas "objects" are just triples connected over a single timestamp.

You might also just get streams of temporal triples in any order (and need to be ordered at arrival only).  

However, let me note that, this is only considering timestamped temporal data now, but not various other forms of *representing* temporal data, where timestampted data is just one of these. As already discussed on the wiki, others are intervals (e.g. in the context of RDF, [1]) but also - which seems not yet discussed - modeling temporal data by change (start/end of existence as "events" (not to be confused with event as before) or "updates", cf. e.g. [2] or even one could think of using sparql11-update sequences as description of streaming data could be viewed as that).

I am not saying that we should cater for all of these options, but probably these need to be discussed/kept in mind.

Apologies if I repeat here discussions that have already taken place, I'd appreciate any hints on that, I am just catching up with the wiki at the moment...

Axel


1. Antoine Zimmermann, Nuno Lopes, Axel Polleres, Umberto Straccia: A general framework for representing, reasoning and querying with annotated Semantic Web data. J. Web Sem. 11: 72-95 (2012)
2. Yannis Tzitzikas, Yannis Theoharis, Dimitris Andreou: On Storage Policies for Semantic Web Repositories That Support Versioning. ESWC 2008: 705-719

--
Prof. Dr. Axel Polleres
Institute for Information Business, WU Vienna
url: http://www.polleres.net/  twitter: @AxelPolleres

On Nov 22, 2013, at 12:55 PM, Axel Polleres <axel@polleres.net> wrote:

> FWIW,
> 
> On Nov 22, 2013, at 12:34 PM, "Gray, Alasdair J G" <A.J.G.Gray@hw.ac.uk> wrote:
> 
>> I completely agree with this. Indeed we should see if we can just repurpose the existing stream definitions for use in the group.
>> 
>>> What RDF gives you is a general framework and a nice general data model, e.g., a graph which changes its shape and constituting data over time and RDF as a nice formalism to capture this.
>> 
>> This is a really interesting point. In the relational model, a tuple carries a certain amount of information which is more than can be captured in a single triple. My question is, should a timestamp be applied at a triple level – with several triples purporting to the same event being given different timestamps – or to a graph – where all triples about an event are captured together?
> 
> I agree that this is a crucial point: In the context of RDF and it's inherent idea to be able to model incomplete data, it is not entirely clear as in relational data when a complete object or tuple has arrived over a stream (which may be viewed as an event per se). I guess this is what makes it difficult to make assumptions about (or even define) things like window sizes, etc. in streamed RDF.
> 
> best,
> Axel
> 
> 
> 
> --
> Prof. Dr. Axel Polleres
> Institute for Information Business, WU Vienna
> url: http://www.polleres.net/  twitter: @AxelPolleres
>

Received on Friday, 22 November 2013 12:28:24 UTC