Permissions and Obligations Expression Working Group Teleconference
12 June 2017
<renato> https://www.w3.org/2017/06/05-poe-minutes
<renato> https://www.w3.org/2017/05/29-poe-minutes
renato: these are minutes from the week before and last week's chat
renato: approved
Resolved: Minutes of 29 May meeting approved
sabrina: alsoRequires tried to come up with an example (horzontal) but don't see the need jus tnow
renato: so we can let it go for now
simon: mandatory uid - we don't allow blank nodes for assets and parties. Why are we limiting ourselves?
simon: if we allow a blank node we can use other properties; e.g. foaf:name
… so we coudl change "must" to "should" perhaps?
renato: party and asset are outside the policy, so we believed that they must be identified by id
<renato> http://w3c.github.io/poe/model/#function
renato: so we recommended a uid + whatever other properties
ivan: agrees it should be a "should". There are cases where you can rely on external tools (owl or rule languages that will deduce correctly). The extra uid will be superfluous
benws: were we using blank nodes when we put constraints on assets and parties?
… we coudl say must be identified by eother a uid or a blank node
simonstey: its different to say we must have an assigner or an assignee. Thats different to saying we must have a uid
… its all expressed in triples. a blank node is different ot a null value
michaelS: uses the example of a webpage. It has an identifier, but the person on the web page has only a name and maybe contact details
<ivan> +1 to michaelS
simonstey: Even if you have a uid you can't always uniquely identify someone
renato: you can declare 2 uids sameAs
ivan: if you can have many identifiers then its not a "must". You can use those and not create a new uid
benws: Theres a distinction between uniquely identifies and identifies uniquely
… the second suggest sit is the only id to use
<ivan> +1 to benws
benws: we need things identified properly but we don't have to insist how it is done
… people must give a party or asset an id which uniquely identifies the thing
benws: but it can be a uid or some other property
renato: so if we change to "should", we need to add that the thing still needs to be uniquely identifiable
simonstey: we should use "should" and recommend strongly that people ought to uniquely identify assets or parties
<phila> PROPOSED: That for asset and party, we change the MUST for uids from MUST to SHOULD, adding a narrative that if you don't use UIDs, then the system needs its own method of identifying the asset or party
Proposal: cFor asset and party we change "must" for uid to "should" and add narrative to say that its recommended to find a way to uniquely identify
<phila> s+//PROPOSED: That for asset and party, we change the MUST for uids from MUST to SHOULD, adding a narrative that if you don't use UIDs, then the system needs its own method of identifying the asset or party
<simonstey> +1
<Sabrina> +1
<phila> +1
<michaelS> +1
+1
<Brian_Ulicny> +1
<renato> +0.9
<phila> (I think Ivan was clearly a +1)
Resolved: That for asset and party, we change the MUST for uids from MUST to SHOULD, adding a narrative that if you don't use UIDs, then the system needs its own method of identifying the asset or party
<renato> -1
<simonstey> -1
<michaelS> bad news
<benws> that would/will be a great loss
phila: has been offered another job and is leaving in three weeks!!!
… ivan is a strong possibility to take over
<simonstey> https://www.w3.org/2016/poe/wiki/Policy_Inference
simonstey: On the Wiki, I've highlighted the what I think the semantics should be
… it will become part of the formal sematics doc (in a more formal way)
phila: I wanted a black box to have a limited function. But group wanted an evaluator to tell you if a constraint is in effect
simonstey: see the outcome of an ODRL evaluator to be to tell you if all rules are in effect
… need to be really careful about what is overriden when there is a conflict
renato: so, do we need a test case?
simonstey: e.g. a policy with permission to present part a with a prohibition to stop Bob from printing part b
… and further W3C prohibition. So outcome could be the rules which are in effect (e.g. if you are Susan)
renato: We have a raw policy, do we expect the evaluator to return a list of rules in effect
simonstey: it would be the expanded policy
phila: then the evaluator needs to know a lot. How do we define how it communicates with the black boxes?
<simonstey> https://www.w3.org/2016/poe/wiki/Policy_Inference#odrl:memberOf
phila: you would need to know the individual, if they are a member of staff at Thomson Reuters. What is the boundary of the evaluator?
simonstey: we ask a black box about a duty and we ask a black box about a constraint
benws: clarification - becasue we have extended relations, the evaluator would need to be aware
simonstey: yes
renato: woudl phil and simon work onthe test cases before Phil disappears :(
simonstey: that will be hard
phila: I can write freestanding text to set out what an evaluator does
Action: phila to draft text defining an ODRL Evaluator
<trackbot> Created ACTION-45 - Draft text defining an odrl evaluator [on Phil Archer - due 2017-06-19].
renato: want ot set a deadline
simonstey: I think it's not only up to the reviewers to raise issues. we can all do that, read through the docs and raise issues etc.
… I think there may be more from me
simonstey: I encourage everyone to at least create a GH account and the raise issues
benws: Where there are sub editing issues, what's the best way of pushing them forward?
renato: Go ahead and do it
renato: It can be more effort to write the issue than to actually make the correction
michaelS: Who will do the make action?>
renato: The vocab - you need Raptor installed to do that.
… If you edit the turtle file - that's the master - you don't touch anything but the turtle file
renato: If I see a change, I can do a make
<benws> I too have to leave
<benws> Congratulations to Phil - but will miss your influence and help
[General discussion about review timing, deadlines etc.]
[None]