- From: Phil Archer <phila@w3.org>
- Date: Fri, 31 Jan 2014 18:18:02 +0000
- To: ashok.malhotra@oracle.com, public-propertygraphs@w3.org
Ashok, everyone, Thanks very much for answering my questions on Wednesday and for preparing the draft charter. I'm exploring options internally and I have two possible routes, neither of which area smooth ride, I'm afraid. 1. We can put the charter forward to W3M as is and try and push it through. We'd need commitment from IBM and other members for this to succeed, otherwise it's borderline (it's all about support and momentum I'm afraid). 2. An alternative is to expand the pool of potential support by expanding the scope. I had an e-mail exchange with Harry Halpin on the topic and he's suggesting that since PGs come from the world of what marketing departments love to call big data, there might be more we could do. How would you feel about a workshop towards the end of the year looking at common data formats including graph data and NoSQL (i.e. JSON data in DBs, ala CouchDB, Mongo, etc.)? Such a workshop could potentially be collocated with TPAC (Santa Clara, end of October). That would introduce a delay which I know will be frustrating as you're hoping to get going soon but it should strengthen the overall effort and, I hope, increase impact. WDYT? Phil. On 30/01/2014 20:36, Ashok Malhotra wrote: > Thanks! I made a couple of changes. > Note sure about the others. We can discuss next week. > All the best, Ashok > On 1/30/2014 1:41 PM, Gregg Kellogg wrote: >> On Jan 30, 2014, at 6:40 AM, Ashok Malhotra >> <ashok.malhotra@oracle.com> wrote: >> >>> I took a stab at a proposed charter for the Property Graphs WG: >>> https://www.w3.org/community/propertygraphs/wiki/Recommendation >> In the Data Model section, you say that nodes and edges can both have >> properties, but later talk about datatypes of attributes. Clearly, >> nomenclature is in-scope for the data model, we should settle on one >> or the other. Most of the references I've looked at relate nodes >> together using "links" or "relationships" Either a node, or a link can >> have properties, which are key/value pairs, where the value seems to >> be a singular scalar value, generally equivalent to an RDF Literal. I >> believe nodes represent entities, which are distinct from literals. >> >> I believe that we would recommend that nodes be denoted by IRIs, but >> allow for unnamed nodes, similar to RDF. Although the keys of >> properties would seem to be strings, it would also be best practice to >> be able to map these keys to IRIs, in a manner similar to JSON-LD, by >> providing an explicit or implicit mapping from keys to IRIs. >> >> Liinks have some notion of unique identifier, which may be entirely >> internal and not expressed in the data model, in addition to a name >> and possibly other key/value paris. We might use the same mechanism >> for transforming a link name into an IRI as for property keys. >> >> I do think that we should consider multi-valued properties, and >> possibly ordered links. Multi-valued properties could be ordered or >> unordered and may or may not allow for repetition; this generally >> corresponds to RDF Seq/Collection/List, Bag and Set, although it >> doesn't seem that there's any provision for shared node and literal >> collections. >> >> As one of the optional deliverables, rather than a simple RDF >> extension, we should more broadly look at the relationship between RDF >> and Property Graphs, how they might be mutually represented, and >> suggestions for future work. >> >> Gregg >> >>> Please take a look. I tried to capture the timeline Phil suggested on >>> Tuesday's telcon but may not have got it quite right. >>> -- >>> All the best, Ashok >>> >> > > -- Phil Archer W3C Data Activity Lead http://www.w3.org/2013/data/ http://philarcher.org +44 (0)7887 767755 @philarcher1
Received on Friday, 31 January 2014 18:18:14 UTC