Notes from irc chat with rdf-calendar group

• the irc log: http://ilrt.org/discovery/chatlogs/rdfig/2003-06-11.html#T16-00-49
• attending noted here: http://rdfig.xmlhack.com/2003/06/11/2003-06-11.html
  • note: the scratchpad is a neat idea, you mark your irc comment with metadata and it gets logged on the scratchpad
• some notes here, because osaf folks were interested and irc logs can be hard to read

Before I got to the chat

• concern that we're inventing new terminology, with suggestions to be found here: BuildOrBuyTerms
• confusion about where to get the details on latest design (hopefully we'll address that with project pages)
• interest in using chandler and haystack as a use case for inference/rule engines where our schemas differ
• understanding that Chandler isn't using RDF for internal representation, but perhaps as exchange format

Notes from the scheduled chat

• interest in p2p sharing of data, although we didn't talk about it in detail
• questions about authorization (permissions)
• explained items as similar to sets of triples, with implementation and performance implications
  • They seemed to understand the tradeoffs, and why we would like items
• questions about information about the same "thing" (using the english word) in different repositories: one item or two?
  • You can think of a triple as coming from some place: (context, subject, pred, object).
  • In practice, you need to keep track of where the data comes from.
  • Need to think about this more carefully.
• general discussion about extensible schema, data flexibility for end users vs. object frameworks for programmers
  • Interesting question: "It seems there are three competeing architectures. 1) Use OO design, and serialize in RDF. 2) Use RDF design, and build objects (items) at runtime; 3) Use RDF design, and use RDF API at runtime instead of OO store. Any feeling which one Channdler is looking like?"
• suggestion: the item store be defined to be equivalent to an rdf store, and test it by making an rdf api to it (in addition to our api). Also, see if the item store can be implemented on top of a general rdf store as an exercise. These exercises would surface differences between our items and rdf. They might be willing to do some of that work.
• they liked the "item can have multiple kinds" feature, asked about it without me mentioning it first
• some comments about zodb, worry that you can't get some of the flexible data features, understand that it is an appealing starting point, lots of good work
• mentioned 'attribute-centric designs': a class (or kind) doesn't dictate an exhaustive list of the properties/attributes its members can have
• cyc ontology was mentioned (example: people can have startTimes and endTimes (birthdays and deaths))
• interest in playing around with Chandler, any early live db/repository
• discussion of uris vs uids, interest in items in Chandler's repository having http uris.
• interest in a view abstraction (encoding views as data, not in python), XUL mentioned as an example
• interest in the query language (or query abstraction)
• CAP mentioned, frustration expressed at separate protocols for separate applications.
• http mentioned as "universal" protocol, something on top to post RDF diffs to an RDF-mapped store.

-- KatieCappsParlante - 21 Jun 2003