- From: Daniele Dell'Aglio <dani.dellaglio@gmail.com>
- Date: Wed, 17 Jun 2015 14:21:29 +0000
- To: Tara Athan <taraathan@gmail.com>, "public-rsp@w3.org" <public-rsp@w3.org>
- Message-ID: <CA+FunMsSKeFCsRcQocyJmNKVMPPdX9wcAbv=m_nvQcVONFtdAQ@mail.gmail.com>
Hi Tara, it seems I misunderstood the meaning of g, i thought it was an identifier of the graph (a URI, or a bnode, as you pointed out). Now I better understand your previous mail. Thanks for the clarification. Daniele Il giorno mer 17 giu 2015 alle ore 14:10 Tara Athan <taraathan@gmail.com> ha scritto: > On 6/15/15 10:28 AM, Daniele Dell'Aglio wrote: > > Hi Tara, > > Regarding this: > > Elsewhere in the document, "triple" appears to be used as meaning "RDF > triple". But in this case "triple" must mean > > simply a tuple of size three, because RDF does not allow a graph to be > the subject of an RDF triple. > > Given that a graph is identified with a URI, it is possible to predicate > on it, as in [1], example 10 (lines 27--29). It follows that (g, p, t) can > be a RDF statement... do you see any problem with that? > > Daniele > > [1] http://www.w3.org/TR/rdf11-primer/#section-trig > > Hi Daniele > > First, I question this assumption: "Given that a graph is identified with > a URI". I don't think every time-stamped graph should be required to have > the graph identified with an IRI, as in some cases, a blank node is > sufficient. Requiring an additional IRI when it is not necessary puts a > burden on fast transmission. > > Second, you are using 'g' as if it is an IRI but the semantics document > says that 'g' *is* a graph. Let's be precise. To make an RDF triple, the > subject must be an IRI or a blank node. It is possible to construct a > triple (n, p, t) where "n" is an IRI or blank node that is the name of a > named RDF graph pair (n, g). Such an RDF triple is not by itself an RDF > Dataset, because it is also required to associate the name with the graph. > However, this RDF triple could be the complete contents of the default > graph of an RDF Dataset d, where d also contains one named graph (n, g). > > It is mathematically sound to talk about a tuple (g, p, t) where g is > actually a graph, rather than the name of a graph. This structure can be in > correspondence to an RDF Dataset, as above. But this tuple cannot itself be > called an RDF Triple. > > Now if the intention is really that (g, p, t) is to be an RDF Triple, then > g *denotes* a graph, rather than *being* a graph -- or it could denote a > name-graph pair, depending on which variant of the semantics > http://www.w3.org/TR/rdf11-datasets/#each-named-graph-defines-its-own-context > is adopted. It is conventional to use "n" as the variable in that case > rather than "g". Then the question arises: is a time-stamped graph really > just this triple? Or is it an RDF Dataset where this triple is in the > default graph? In the former case, then the IRI must be dereferenced or > otherwise resolved to obtain the graph for querying. In the latter case, > the named graph must be included in the stream in order to fully define the > RDF Dataset. > > Finally, the RDF Primer is not intended to ever be normative. As such, it > can contain statements that are not entirely precise, because it is about > introducing newcomers to the ideas, not defining them. The TRIG > specification (http://www.w3.org/TR/2014/REC-trig-20140225/) is a > normative specification that defines a concrete syntax for RDF Datasets, > and (rightly) says nothing about the semantics of RDF Datasets. But when we > are defining the abstract model for a time-stamped graph, we should not be > referring to any particular concrete syntax. The RDF Dataset semantics is > not yet an accepted recommendation, but it is the document most likely to > progress to a recommendation, and I think it is the best choice for the > foundation of time-stamped graph semantics. > > This whole discussion is relevant to the issue > https://github.com/streamreasoning/RSP-QL/issues/10, but perhaps > qualifies for an issue of its own. > > > Tara > > > > > Il giorno lun 15 giu 2015 alle ore 15:16 Tara Athan <taraathan@gmail.com> > ha scritto: > >> On 6/15/15 5:30 AM, Gray, Alasdair J G wrote: >> >> Hi Tara, >> >> On 14 June 2015 at 12:00:09, Tara Athan (taraathan@gmail.com) wrote: >> >> Dear Abraham, and all - >> Please excuse me if this point has already been discussed in the group, >> as I am late joining the discussion. >> >> Welcome to the discussion, the more the merrier. >> >> >> It seems to me that there is an existing basis on which to build such a >> data model - the RDF 1.1 dataset. The semantics for a set of time-stamped >> graphs (g_i, p_i, t_i) that seems most appropriate to me is the one defined >> here: >> >> http://www.w3.org/TR/rdf11-datasets/#each-named-graph-defines-its-own-context >> and the name of each graph would be an implicit blank node that is also >> the subject of a triple in the default graph. This triple has predicate p_i >> and object t_i . >> >> The discussion of the streaming graph data model came up at our recent >> face-to-face meeting which is where we came up with the current data model >> described in >> >> https://github.com/streamreasoning/RSP-QL/blob/master/Semantics.md >> >> As you will see in that document, we have exactly the semantics you are >> suggesting. >> >> I see this now - in the section "Timestamped Graph" - it is somewhat >> hidden as RDF Datasets are not explicitly mentioned here. There is one >> point especially about this definition that I find confusing: '(g, p, t)' >> is called a triple. >> >> Elsewhere in the document, "triple" appears to be used as meaning "RDF >> triple". But in this case "triple" must mean simply a tuple of size three, >> because RDF does not allow a graph to be the subject of an RDF triple. >> >> I have put this as well as a few other clarifications into a pull request >> (https://github.com/streamreasoning/RSP-QL/pull/12) for purpose of >> discussion. I noticed that the pull request is already merged - I probably >> should have made clear in my pull request comment that this was requested >> for discussion, rather than immediate merge. >> >> Best regards, Tara >> >> Alasdair >> >> >> >> Tara >> >> On 6/14/15 3:59 AM, Abraham Bernstein wrote: >> >> Dear Emanuele, dear all >> >> I wonder whether we are mixing two issues here. One is the data model >> of time-annotated graphs. The other is a system model that, as you >> indicate, is much easier to deine if you can make some assumptions about >> how the triples (or graph fragments) arrive (in order, monotonically >> increasing, etc.). >> >> I would propose to disentangle the two. In other words, I would propose >> a well-founded time-based data model combined with a set of assertions that >> we expect to hold on streams. >> >> Best >> >> Avi >> >> >> >> On 12.06.2015, at 18:16, Emanuele Della Valle < >> emanuele.dellavalle@polimi.it> wrote: >> >> Dear Alasdair, >> >> a problem I run into went I implemented the timestamped model in real >> use cases is that you need to wait for all contemporaneous triples with the >> same timestamp, before processing them. They arrive to the RSP engine one >> after each other, so the arrival time is always increasing, but they all >> carry the some timestamp. If you assume that timestamp are not decreasing, >> an RSP engine knows it can start the processing as soon as a triple with a >> larger timestamp arrives, but what if the stream stay silent? How does the >> RSP engine distinguish the case of a delayed triple (still contemporaneous >> to those it has already got) from the case it is waiting because nothing is >> transmitted on the stream? In the C-SPARQL engine we decided to give up >> with the possibility to treat the application time and we only relay on the >> receiving time. This is also what STREAM does. It is know as the best >> effort approach. Esper can work in best effort mode, but you can also send >> an event to say the time is past. This is call external time control. This >> time keeping event is a form of punctuation. It means, I told you all I >> have to say at this point in time. >> >> If graphs are timestamped with a strictly increasing timestamp, then as >> soon as the RSP engine gets the entire graph, it can process it. In other >> words, the boundary of the graph is a form of punctuation. If another graph >> with the same timestamp can follow, than you’re back into the problem you >> cannot distinguish if you are waiting for a delayed graph with the same >> timestamp from the case the stream is silent. >> >> I hope I expressed myself in a clearer way this time. >> >> Best Regards, >> >> Emanuele >> >> PS I’m in favour of multiple time annotations and I agree that >> interval-based semantics matters. >> >> >> >> >> On 12 Jun 2015, at 18:31, Gray, Alasdair J G <A.J.G.Gray@hw.ac.uk> >> wrote: >> >> Dear Emanuele, >> >> I don’t quite follow the punctuation argument meaning that we can only >> have one graph at any given time point. >> (Unfortunately I’m on the train home and cannot access the article that >> you linked.) >> >> We still have the gain over the traditional streaming RDF model in that >> all triples conforming to a given observation will be contained in the >> graph. So why does having more than one graph at a given time point cause a >> problem? >> (Sorry if I am missing something obvious) >> >> Best regards, >> >> Alasdair >> >> On 12 June 2015 at 08:49:40, Emanuele Della Valle ( >> emanuele.dellavalle@polimi.it) wrote: >> >> Dear Alasdair, and all >> >> thanks for the report. I would like to point out that the sentence >> “There can be multiple graphs with the same timestamp” is, in my opinion, a >> bad choice. It will prevent graphs to be interpreted as a form of >> punctuation [1] and this was one of the most important gain of the version >> of RSP Data Model discussed in Berlin (i.e., graphs with strictly >> increasing timestamps). The lack of punctuation is a problem of the >> “traditional" timestamped triples data model where contemporary triples >> must be admitted. >> >> Best regards, >> >> Emanuele >> >> [1] >> http://link.springer.com/referenceworkentry/10.1007%2F978-0-387-39940-9_285 >> >> >> On 11 Jun 2015, at 18:37, Gray, Alasdair J G <A.J.G.Gray@hw.ac.uk> >> wrote: >> >> Hi All, >> >> During the ESWC RSP Workshop we had a breakout group focus on defining >> the RSP data model. I was charged with the action of updating the semantics >> document with the agreed model. >> >> You can find the updated data model at >> https://github.com/streamreasoning/RSP-QL/blob/master/Semantics.md >> >> Best regards, >> >> Alasdair >> >> -- >> Alasdair J G Gray >> Lecturer, Heriot-Watt University >> Web: http://www.alasdairjggray.co.uk >> ORCID: http://orcid.org/0000-0002-5711-4872 >> Twitter: @gray_alasdair >> Telephone: +44 131 451 3429 >> Office: EM 1.39 >> >> >> >> We invite research leaders and ambitious early career researchers to join >> us in leading and driving research in key inter-disciplinary themes. Please >> see www.hw.ac.uk/researchleaders for further information and how to >> apply. >> >> Heriot-Watt University is a Scottish charity registered under charity >> number SC000278. >> >> >> -- >> Alasdair J G Gray >> Lecturer, Heriot-Watt University >> Web: http://www.alasdairjggray.co.uk >> ORCID: http://orcid.org/0000-0002-5711-4872 >> Twitter: @gray_alasdair >> Telephone: +44 131 451 3429 >> Office: EM 1.39 >> >> >> >> We invite research leaders and ambitious early career researchers to join >> us in leading and driving research in key inter-disciplinary themes. Please >> see www.hw.ac.uk/researchleaders for further information and how to >> apply. >> >> Heriot-Watt University is a Scottish charity registered under charity >> number SC000278. >> >> >> >> ----------------------------------------------------------------- >> | Professor Abraham Bernstein, PhD >> | University of Zürich, Department of Informatics >> | web: http://www.ifi.uzh.ch/ddis/bernstein.html >> >> >> >> >> -- >> Alasdair J G Gray >> Lecturer, Heriot-Watt University >> Web: http://www.alasdairjggray.co.uk >> ORCID: http://orcid.org/0000-0002-5711-4872 >> Twitter: @gray_alasdair >> Telephone: +44 131 451 3429 >> Office: EM 1.39 >> >> >> >> We invite research leaders and ambitious early career researchers to >> join us in leading and driving research in key inter-disciplinary themes. >> Please see www.hw.ac.uk/researchleaders for further information and how >> to apply. >> >> Heriot-Watt University is a Scottish charity registered under charity >> number SC000278. >> >> >> >
Received on Wednesday, 17 June 2015 14:22:10 UTC