See also: IRC log
<renato> 12:30 pm | Greenwich Time (Reykjavik, GMT) | 1 hr
hi
i can scribe if you like
<scribe> Scribe: victor
<renato> https://www.w3.org/2017/06/19-poe-minutes
<michaelS> scribenick: victor
<phila> (not present)
RESOLUTION: last meeting's minutes are approved
<benws> hmmm - problems calling in
renato: I have explored the
practices of other groups and their exit criteria
... the w3c mawg listed "features" which were implemented by
implementors. they were the key selling idea.
<renato> https://github.com/w3c/poe/blob/gh-pages/test/cr-exit.md
renato: we can create test cases for each of the features listed in the url above
phila: I had an action item due,
currently almost done trying to identify what an ODRL Evaluator
had to do
... I made notes to the examples, the exit criteria would be to
check whether the output of the evaluator is the one expected
given one or more policies.
... the key is "what an evaluator must do"
renato: constraints in any case
are evaluated by black boxes.
... validation is first
phila: i am trying to write down what happens one the black box has worked, "not how the black box works"
<simonstey> +q
simonstey: is the output yes/no
or the policy containing the set of rules which are in
effect?
... e.g. given a policy with 20 rules, shouldn't the policy
evaluator say which is/are the rules which are in effect?
<phila> victor: Is the ODRLE stateless?
<simonstey> +q
<phila> ... A god feature IMO is that it would be
<phila> simonstey: You don't really have a choice of which rules you can take
<phila> ... If the OE is evaluating a request against a policy, all the rules apply, you can't ignore some
simonstey: for each request, there is as an output
<phila> victor: Yes, I'm saying there should be only one request for the test cases
victor: yes, in a system without memory, the test cases should only have at most "requests" but not "sequences of requests".
simonstey: we cannot foresee how the odrl evaluator should be as long as the test cases are passed
<simonstey> +q
renato: it is quite likely that some implementations only consider validation, whereas "evaluation" will not be regarded.
<benws> Sorry guys - I can't call in. V. odd.
simonstey: the absolute minimum
should abide to the IM in its section 2.7 onward.
... conflict resolution etc. has to be checked/tested
... propagation of constraints in the root level etc.
<phila> +1 to simonstey
<renato> https://github.com/w3c/poe/blob/gh-pages/test/cr-exit.md
simonstey: we need to check whether ODRL implementations do that, unless we use "MAY"s. etc. The absolute minium is "given a policy or set of set of policies, possibly with conflicts and conflict resolution information, reduce the set of rules to the result of applying the spec".
renato: in the web above, you can
see that a feature is "A Policy that includes Policy
inhertiance".
... we can write test cases for each of the bullet points
above
victor: Phila, have you written anything?
phila: hopefully today I'll distribute something
renato: is there anything missing
from the exit criteria list?
... is this ok?
phila: yes
subtopic: who are the potential implementations?
renato: we should contact potential implementors out of the WG
<renato> https://github.com/w3c/poe/blob/gh-pages/test/implementors.md
renato: we should try to fill the
table provided in IRL link above
... Caroline could have feedback from three organizations
(?)
ivan: 3 implementations are in the low side
renato: indeed
+1
<renato> Fraunhofer Gesellschaft
victor: UPM will provide an implementation, as long as abiding to the test cases is not terribly complicated
renato: possibly Fraunhofer will also.
<renato> https://github.com/w3c/poe/issues/164
victor: UPM has staterd in a Pubby fork to serve linked data conditionally.
renato: the MAWG has decided to
use "source" to prevent id hijaacking
... so for the collections, "source" would be used, and not
"uid"
<simonstey> +q
<michaelS> akc m
<simonstey> https://github.com/w3c/poe/issues/201
michaelS: this has to be explained to the users of the information model. source is an alternative to id; but why/how?
simonstey: by not using uid, blank nodes may live. Then, additional constraints identified by a certain URI might be concealed.
renato: it is about creating another property which would be "source".
ivan: not sure if the background
of the problem is clear. in RDF, once a triple is out there in
the web it is not confined, and anyone can see.
... so if someone makes a statement on a resource on the web
("ivan is blonde"), this is public, and this is what hijaacking
is about.
renato: does anybody oppose?
(silence)
<renato> https://github.com/w3c/poe/issues/162
renato: it is about a policy
saying "pay me 5 dollars".
... which as of today, clashes with the current IM definition,
which associates duties only at the permission level.
<simonstey> +q
renato: root-level-duties may prove useful for inheritance in an agreement.
simonstey: i already commented in
the github, there is an inconsistency now: we said that the IM
specifies that a policy must have at least a rule (permission,
prohibition, duty), but then we say that duty can only hang
from a permission.
... if a duty is added to an offer/agreement then there must be
one permission. otherwise, there cannot be a duty alone.
michaels: this is related to
shortcuts at the policy level.
... we have some already shortcuts
victor: I want some clarifications in the shortcuts
<renato> https://w3c.github.io/poe/model/#composition
simonstey: the duty case is
different from other shortcuts like "odrl:mypolicy
odrl:permission odrl:play".
... in the first case, it can be specified otherwise, yes, but
with too much effort
victor: I saw and advantage in having a policy defined by a set of triples in the form "odrl:mypolicy odrl:PROPERTY odrl:VALUE".
renato: we can see them not as
"shortcuts" but as first class objects
... the duties are naturally "top level" in some cases.
simonstey: there are some duties with an implicit permission. consider GDPR. One may say "if you want to handle with personaldata you have to do X"
victor: if we have two manners of expressing something, we can define a CANONICALIZATION operation (much like in the XML world canonicalization is)
michaelS: (explains a case victor does not grasp)
renato: he means: we have
non-conflicting permission A and prohibition B, and a duty at
the policy level.
... what is the meaning of the duty once being satisfied?
simonstey: you only pay for your right, but the prohibition also holds.
renato: we need more discussion on this.
RESOLUTION: to accept "source" for collections as an alternative identification mechanism
renato: please join the github issues
<simonstey> :(
<michaelS> says thanks to phila
ivan: will set up a new W3C call
for the next weeks
... but i will be on vacation for 4 weeks
... ralph will fix the webex meetings'
phila: this is perhaps my last
call
... i will clean up the minutes for the last time.