Telecon 14.11.2014

Participants

Peter

Please also refer to the comments and discussion going on in the RSP Data Model Google Doc.
Mark Feblowitz (Comcast) raised some interesting questions on the mailing list relating to the point in time that queries are executed. See [2] + answers.
Currently there is also ED's example of an RSP-QL syntax under discussion. Please refer to [3] and [4].

[discussing the comments of the Google Doc]

ED: It's important to discuss the issue of blank nodes. I wrote about it in the google doc. I just want to point it out and discuss it later.
OC: I would go for as doing it the same as with the named graphs in RDF.
ED: We should review it, and if it makes sense, we should do it like that. If we dont have a graph, and only the blank nodes in the window, it can cause some damage.
OC: I will write an e-mail after reviewing this issue.
ACTION (OC): Write an e-mail after reviewing the issue of how to handle blank nodes.

[Question arises: What is the output of a query?]

ED: I reported the design decision of C-SPARQL. We broke down to be compatible to the input. And restrict some of the query forms.
AG: There are still a lot of open questions there.

[discussion about s2r, r2r, etc. windows]

DL: Windows should match a triple pattern, not a single triple. In CQELS I count the amount of matched patterns.
ED: +1
OC: Danh, do you count BGPs?
DL: Yes
AG: And your BGP can be a complex BGP, can it?
DL: Yes, i think so. But it should be done more formal.
AG: DL, can you write up the proposal in the wiki so we can discuss it?
ACTION (DL): DL to write up his proposal for BGP count windows on the wiki.
DL: Yes, will do.

AG: Question: When should you evaluate a window. At the end of a window, or by the time of arrival of triples/tuples?
ED: Should we let the user decide? Should we put it in the semantics to be able to control it? Do we want to be as complex as EPL?
AG: Since we don't have too many users, we should consider complexity of our language a key issue.

AG: We should start a process to agree on a syntax on the wiki. A syntax for registering streams, for doing windows, for representing these graphs as streams.
ED: The language will only allow to access the identifier of a stream. Defining the semantics first will kill us.
OC: Yes, I think the syntax is the best way to somehow encode semantics later on.
ED: I can write a query which is a hybrid of all languages. A prototype of a query of the new language.
DL: +1
ED: We should do it use case based. And if we had enough discussion, we can do it formal. I will try to stick to the room example.
ACTION (ED): ED will create this hybrid query in the wiki and then put it in a Google doc (siemens guys cannot access google docs).
AG: Let's close. The agenda for next time is: discuss syntax that comes out of ED's document.
ED: I will not be available next week and one week after. I propose 27th of November
AG: Okay, I will propose it via mail and see if there is a veto for 27th.

ACTION (OC): Write an e-mail after reviewing the issue of how to handle blank nodes.
ACTION (all) discuss the issues regarding s2r, r2r, etc.
ACTION (DL): DL to write up his proposal for BGP count windows on the wiki.
ACTION (ED): ED will create this hybrid query in wiki and then put it in google doc (siemens guys cannot access google docs).
ACTION (all): comment on hybrid RSP-query from ED until 27th of November (Thursday!). see: [5] and [6]