Minutes from Jan 14 Property Graphs CG Telcon

No Telcon Next Week.
Next Telcon Jan 28, 2014

[11:57] <Ashok> meeting:Property Graphs CG
[11:57] <Ashok> Chair:Ashok
[11:58] <Ashok> Regrets:Andy, Michael
[12:03] <Ashok> present: Gregg, Kelvin, Ted, Zhe, Ashok
[12:04] <zwu2> scribenick: zwu2
[12:04] <zwu2> Topic: review last week's meeting minutes
[12:04] <Ashok> - Approval of minutes from last meeting:http://lists.w3.org/Archives/Public/public-propertygraphs/2014Jan/0025.html
[12:05] <zwu2> Approved minutes from last meeting:http://lists.w3.org/Archives/Public/public-propertygraphs/2014Jan/0025.html
[12:06] <zwu2> Ashok: next telecon will be on Jan 28th
[12:06] <zwu2> ... we will skip one week
[12:07] <zwu2> ... we will have Phil Archer from W3C with us
[12:08] <zwu2> ... no new member info this week
[12:08] <zwu2> ... no action on use cases
[12:09] <zwu2> ... want to discuss what we want this CG to recommend
[12:09] <zwu2> ... to the W3C
[12:09] <zwu2> ... PG as a data model
[12:10] <zwu2> Kelvin, like to see the proposal written down, what is the focus point
[12:10] <zwu2> ... should be some concrete things that we think we need to work on
[12:11] <zwu2> ... how it will work with what is going on in the industry
[12:11] <zwu2> Ashok: at least let us start speaking about it
[12:11] <zwu2> ... my personal view: we would agree on an abstract PG model
[12:12] <zwu2> ... we have questions on data type attributes
[12:12] <zwu2> ... there are some emails recommending xsd datatypes
[12:12] <zwu2> ... a further question, whether we want collection types
[12:13] <zwu2> ... Andy recommends this
[12:13] <zwu2> ... graphs/attributes/datatypes are all at an abstract level
[12:14] <zwu2> ... I think this abstract model can be serialized into multiple formats
[12:15] <Ashok> zhe:Good to start with abstract datatypes
[12:16] <zwu2> Ashok: yes we are going to recommend an abstract data model
[12:16] <zwu2> ... collection type,
[12:16] <zwu2> ... and its representation
[12:17] <zwu2> Kelvin, there are more domain specific things that are more interesting
[12:17] <zwu2> ... schema.org,
[12:17] <zwu2> ... I don't see the abstract data model as the biggest problem
[12:17] <zwu2> ... we should continue our discussion
[12:18] <zwu2> Ashok: regarding schema.org, are you saying we should use attributes recommended by schema.org?
[12:18] <zwu2> Kelvin, we will all probably build graphs for customer applications
[12:19] <zwu2> ... especially in the social graphs area
[12:19] <zwu2> gkellogg: schema.org is not the XML schema for XML
[12:19] <zwu2> ... it is a vocabulary
[12:19] <zwu2> ... entities/properties/classes denoted by IRIs
[12:20] <zwu2> ... does this group recommend the use of IRIs for property graphs,
[12:20] <zwu2> ... my understanding is that PG does not use it
[12:20] <zwu2> ... attributes are local to the graph
[12:21] <zwu2> q+
[12:22] <zwu2> Ashok: if you were to do a PG data model for a domain
[12:22] <zwu2> ... you can use attribute names as IRIs defined for domains
[12:23] <zwu2> Kelvin: linked property graphs
[12:24] <TallTed> general class == property graphs
[12:24] <TallTed> subclass == linked property graphs (using IRIs) <-- a/k/a best practice :-)
[12:26] <zwu2> TallTed: need a firm model for PG
[12:26] <zwu2> ... then linked PG
[12:27] <zwu2> Ashok: we ought to create a PG model that allows use of concepts/IRIs defined elsewhere
[12:28] <zwu2> Topic: APIs
[12:28] <zwu2> Ashok: I thought we ought to do a data model and APIs
[12:29] <zwu2> ... in the last few telecons, got some push back
[12:29] <zwu2> ... 1) Andy said we should scope it down
[12:29] <zwu2> ... so that we can finish in time
[12:30] <zwu2> ... 2) and, REST based APIs do not work well, don't scale, etc.
[12:30] <zwu2> ... we should think about a declarative one
[12:30] <zwu2> ... like a query style API
[12:31] <zwu2> gkellog: the most important thing this CG can do is a data model
[12:31] <zwu2> ... the APIs are premature
[12:31] <zwu2> ... would be useful to have a serialization
[12:31] <zwu2> ... not to claim that there is a single serialization,
[12:31] <zwu2> ... perhaps something like N-TRIPLES, N-QUADs
[12:32] <zwu2> ... if this CG settles on that, it will lay a great foundation
[12:34] <zwu2> Ashok: data model + one simple serialization
[12:37] <Ashok> Kelvin:What problem does the serialization solve
[12:37] <Ashok> Zhe:Default interchange format ... serialization to test model
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[12:40] <zwu2> Ashok: Kelvin, you want a simple, readable format
[12:40] <zwu2> ... we also need a highly scalable format ... these are different usecases
[12:42] <zwu2> ... I will try to put something up on the wiki
[12:43] <zwu2> ... feel free to edit
[12:43] <zwu2> ... also feel free to start an outline
[12:44] <zwu2> Topic: any other business?
[12:44] <zwu2> ... see you in two weeks
-- 
All the best, Ashok