See also: IRC log
<phila> scribeNick: phila
phila: Will start
Serena: Second slot
michaelS: Later, after constraints
<simonstey> simon: I'll do the 3rd one
<michaelS> I'll do the 4th
renato: I'd like to go through
approved req issues (green ones) first
... The go through editorial ones after that
... Going from GH issues
<renato> try this: https://mit.webex.com/mit/j.php?MTID=me949e9479628ccd68d29fec3cc71c0da
<michaelS> +1
benws: Any objection to https://www.w3.org/2016/11/28-poe-minutes.html?
<Serena> +1
RESOLUTION: Minutes of 28 Nov approved
benws: See https://github.com/w3c/poe/issues
<renato> https://github.com/w3c/poe/milestone/1
-> https://github.com/w3c/poe/issues/62 Issue 62
scribe: See also https://github.com/w3c/poe/issues/59
... 58
... 56
... 60
<michaelS> please add 81
renato: We've seen examples on the list and we've discussed this a lot in the telcos. Need to come to a consensus around constraints on constraints
<renato> https://github.com/w3c/poe/issues/62
renato: There are various
comments on that
... including some from michaelS
... Questioning whether we need constraints on constraints
michaelS: From my reading -= what
exactly is a constrainty on a constraint?
... In some cases it's combining constraints. But I think it's
altready defined that multiple constraints are ANDed
... so what is a constraint on a constraint?
renato: The idea is that a c on a
c, they still both apply to the action
... But the 2nd constrraint applies to the first
... Classic example was the 30 mins after a football match
image
... So action is to distribute the photo
benws: It's a technique. It's
where there is more than one right operand to check
... Constraints allow us to do that
renato: That's a fair
description
... So a ConC helps to resolve issue 60
... If we had the ability to say what the actual subject of the
constraint is explicitly, then we can say that a C is
constyraining another C.
... Issue 60 then helps us solve issue 62
... The solution is, in the model, we have another attribute of
subject. That allows us to explicitly say that the constraint
is on the action, or it can be on another constraint
benws: Given that a constraint has a name, can't we... [breaking up]
<simonstey> +q
benws: ... it has a URI, so we can address it ... can't we use odrl:constraint to attach the constraint in the same was we do something else
simonstey: If we start to attach
ConC, what are the intended semantics?
... e.g. permission, a C might give a period of time. If attach
an additional C, what does that mean>? Can i attach any kind
of C to any other kind?
... What if there is a large chain?
... What does an arbitrary set mean?
michaelS: My question is in a
similar direction.
... IPTC has RightsML use cases that incluide combined
constraints. Like image can be used everywhere except UK but
only for a limited time
... But we do that already
<simonstey> +q
michaelS: I don't see a need to say that the temporal constraint is a C on the spatial one
benws: In that example, you're correct. But an embargo can't
<michaelS> http://dev.iptc.org/RightsML-Combined-Example-geographic-and-time-period
<simonstey> -q
benws: Each right operand can be
checked separately, but for an embargo, you need to know both
end of the event and the hour after
... An eg I've put in the constraint renaming paper. I've put
an example to use this left operand resource, which identifies
the event. And the C defines a period after the event
... And that's all in a single constraint. So you have 2 right
operands in a single constraint?
michaelS: No, I added a resource
to the left operand
... So the end event is the left operand
... and the right operand is the period afterwhich the action
can be performed
benws: OK, that might work for an
embargo
... But there are other more complex ones
michaelS: We wanted the flexibility of the free text
asck me
<Zakim> phila, you wanted to womnder aloud if we ever need more than 2. If not, can we think of sub constraints
<michaelS> q
victor: Is the name ConC a problem? How about extended info on constraints? Or is it more than that?
simonstey: I think it's way more
than that.
... Even this sub constraint, what does that mean? It means
different things. e.g. for a pay amount of a daytime
constraint
... Inheritance comes in. Are things overwritten or partially
overwritten?
... But I think it would help to not allow an arbitrary
collection.
... Extending the left operand might work.
... It's those extended relations that might hold the
key.
... Some ORing, maybe
... Another option might be to constraint a whole package
victor: AIUI, this idea of ConC
operates differentely in different examples
... The only common point that we have is that it provides
extra info and a simple constraint can't be defined
... I agree that a clear model is more important than a clear
name.
simonstey: It's important to
state that the time period only begins when X happens
... There are situations where you don't want them both checked
at the same time or sequentially.
renato: From this conversation -
do we have enbough info that we can make a sensible
decision?
... It sounds as if there are odd cases where it works but the
general is more problematic
... If we can't come up with a good solution maybe we shouldn't
keep trying
benws: We're doing it with a
payment constraint
... The additional unit, GBP, USD
... If we add in the unit of count, we're sneaking in an
additional predicate.
... We are creating a a special case. You sometimes need an
extra predicate
victor: Do you have an example online?
benws: If you have a payment amount, you need to include the currency
<renato> https://lists.w3.org/Archives/Public/public-poe-wg/2016Nov/0065.html
benws: If it's about location, you'll need to include IP address or whatever
renato: That example has a pay amount example with a unit and a unit of count
michaelS: I think we should make
a distinction... the right operand is an entity and that could
have properties.
... If you want to express a financial amount, that includes a
currency, it's not a constraint, just a property of the right
operand
... Issue 58... we talked about what a unit of count actually
is. I feel it's another constraint, not a ConC
benws: But if you define the amount that needs to be paid, in order to define the unit, how do you connect that to the pay mount comnstraiont?
michaelS: The unit is part of this constraint.
simonstey: I agree with Ben on
that issue.
... There are situations where ConC makes sense, e.g. pay 100,
then say the currency.
... So that's a constraint on a constraint.
... This combination makes sense, and we've sneaked this into
the model
... But there other situations where this unit thing doesn't
make sense.
... it's saying you can't pay $100, you have to pay €100
michaelS: I can't agree with
that. If you want to define an amount for payment... this
colour value has no reason.
... If you have to say how much money, you need to say what
currency
simonstey: If I don't care about the unit, it's not wrong...
benws: It's even more obvious in
unit of count. Is it for an access ID, a physical ID etc.
... Otherwise we don't know what the payment is for.
michaelS: In my answer to issue 58, I said why is this not a constraint
benws: Because the subject of the constraint, current model, is the action.
michaelS: But these are siblings
renato: So in
https://lists.w3.org/Archives/Public/public-poe-wg/2016Nov/0065.html
... Look at the duty
... There's a constraint of a unit of count and a time
interval.
... So the currency is constrained and there's a time
constraint
michaelS: Unit is a part of the amount
<Zakim> phila, you wanted to remark that's still only 2 constraints and to talk about https://www.w3.org/TR/dwbp/#LocaleParametersMetadata
-> https://www.w3.org/TR/dwbp/#LocaleParametersMetadata
phila: Talks about locale neutral data and only 2 levels of CofC
benws: On the first
implementation, I might think that 2 levels is enough. Then I
might find a future more complex version that needs a chain and
I'd have to write code to support any number of
constraints
... I see no reason to limit us to 2. 3 would be rate, 4 very
rare, but the world is a big plavce.
phila: OK
renato: So where are we?
benws: I'd like to understand Michael's solution for an embargo more
renato: Did you write some pseudo code for that michaelS?
michaelS: Regarding this event and period, I have put this in the constraints
<michaelS> https://www.w3.org/2016/poe/wiki/Constraints#Definition_of_the_Left_Operands.2FNames_of_Constraints
michaelS: That says left operand
defines the end of the event
... The left operand resource states sthe end of the event. The
right operand states the period. After that the action can be
(not)performed
benws: So if we were to look at a
pay amount constraint
... Would you say that left operand unit: Euro?
... Would that be a solution?
michaelS: We could use it like that. We have our pay amount...
simonstey: You're saying that you
can have those generic left operands and explain in plain text
what those left operands are
... So you end up with a a super-large left operand
... The machine provides the end of the event, but I have to
know that the end of the event is not when this starts.
michaelS: This left hand operand would give the end of the event as the start. You get the 30 mins from the right
benws: Can a machine understand this?
michaelS: We talked about this on 15/11. WE said that most of the processes need to be hand crafted
<Sabrina> this sounds like linear constraints which is outside of OWL
michaelS: You can have lots of left operands, and you have to adjust your code
benws: You might have that, but I'm hoping for profiles that can automate this process.
michaelS: That's why I proposed
to add the left operand so you can see it automated.
... SO maybe TR can define that count has to be taken from a TR
resource and this resource is the reference for checking
against the value of count in the right operand/.
<renato> https://lists.w3.org/Archives/Public/public-poe-wg/2016Nov/0074.html
benws: When you say that, it sounds right, but then I see you have to read the text. It seems the language doesn't have the semantics to express something relatively simple.
renato: I was trying to work out
that the left operand 1-9 meant
... Is there a URI for getting the current count, for
example?
michaelS: Yes
... The free text defines that you have to check a reference
against the right operand. To have that machine processable,
you can provide the URI that gives the official count.
... And Ben you said that free text doesn't make you happy.
What I propose allows using an API for retrieving the count
value
renato: Does you event end period, does that contain a URI of the event or the period of time you wait after the event.
michaelS: The simple solution is
that the left operand provides an ID for the event. But that
doesn't work as you want the end of the event, not the event
itself
... But an event organiser might provide a URL that provides
the timestamp of the end of the event. Ping thsat until you get
a timestamp and then start counting.
renato: We're getting into a
processing model - how a system would communicate
... I'm sending a request to a URI and whether I get a response
and what response I get starts to matter.
benws: It seems we're jumping to the processing model far too early
renato: So where do we go from here on ConC?
<simonstey> can't we just point to an instance of type https://www.w3.org/TR/owl-time/#time:TemporalEntity
<victor> +1
renato: Is there agreement that we need one level of ConC?
<Serena> +1
benws: I think we def need one or more
<Sabrina> +1
renato: A profile can restrict numbers of constraints, from a model POV, we need to allow the general case
michaelS: To clarify... which count are we talking about? Are we talking about the levels of constraints, or how many consytraints at the first level?
renato: In theory, we can have 10
constraints
... The Q then is, within any of those Cs, are they further
constrained, should we put a limit on how many there could
be?
<Serena> *phila, do you want me to take the hand for scribing?*
<Serena> *ok!*
benws: Let's clarify ...
... Whatever solution should be machine processable. Secondly
may require more than 1 predicate/.
PROPOSED: Constraint handling should be machine processable
renato: Expressable or processable?
benws: I want the language to be able to express permissions without having to read free text.
renato: Is 'industry is education' is that machine processable?
benws: Yes
michaelS: The Q is, is this as
strict so that all constraints MUST be machine processable, but
is it a goal?
... I only want to point to the existing constraint that says
you need to read the policy
<simonstey> +q
benws: Include those in scope
ivan: I'm a little bit scare of this... if we do that, then we will have to prove interop. Then we get into a required processing way.
benws: I don't think we're
specifying a means of processing it [Breaking up]
... [missed] in theory processable
ivan: We need to be careful how we formulate that
benws: We're not specifying a processing model, at this stage, we're saying at this stage that a machines doesn't need to go off and read free text.
ivan: In theory I agree with
that. But if we have constraints that include time
... So we'd need to specify this
benws: I think Michael has done a lot on this
ivan: Isn't it better to say that
ODRL should be self contained and not depend on additional
prose.
... I am worried about the terms processing and processable
here. It raises potential flags and interop issues
simonstey: The Q is whether you
meant verifiable
... There are things you can't necessarily check
<ivan> Proposed: ODRL constraints SHOULD be self contained and not depend on additional prose.
simonstey: Something like 'review policy'
ivan: That'[s why it says SHOULD
not MUST
... There might be cases where this is not achievable, but the
goal is not to depend on additional propose
<Zakim> phila, you wanted to talk about related current future work
phila: Talked about verifiable claims and OWL Time
michaelS: Does ODRL have to
express *everything*?
... A typical use is to complement something like a 10 page
contract
... for example, constraints of spatial and temporal use. These
are typical in news.
... The machine readable data will need to interpret the full
text.
... There may be constraints in a 10 page contract that are not
expressed in the ODRL policy
... So I'm not happy about having a strict requirement that
everything must be machine readable
benws: I didn't mean that
michaelS: That's how I read it.
benws: It's clear to me that an
ODRL policy is an interpretation for a purpose
... It should be understood without ref to free text
victor: Or at least say that for a full interpretation you need to see the full text
michaelS: We have a duty for that
benws: There's no sense on my part that ODRL is meant to be a full representation of the contract. That won't happen
victor: Can we matrk those constraints that can and can't be digitally evaluated?
benws: Let's come back to that. And come back to the proposal
victor: I think it's the
same
... distinguishing those that are self-contained and not
benws: I think you're conflating expressability and processability
Proposed: ODRL constraints SHOULD
be self contained and not depend on additional prose.
... ODRL constraints SHOULD be self contained and not depend on
additional prose, noting that it may be a subset of a larger
human readable policy.
<CarolineB> +1
<ivan> +1
<Sabrina> +1
<Serena> +1
<michaelS> 0
<renato> +0.5
<simonstey> +1
<michaelS> What exactly is "additional prose"?
michaelS: I put my user hat on. If you want to process a constraint, you have to read the left operand. Is that dependent on additional prose?
benws: No
... you have to learn the language
simonstey: I think michaelS is talking about custiome left operands
<victor> +0.5 (explanation: some constraints should be self-contained and processable, some others can be "additionalTextConstraint". The important thing is to have ANNOTATED the status of each term)
michaelS: So again for this count value.. it doesn't say from where you get the counts.
benws: No, it means that the
language should be riche enough to tell people where to get the
count from
... How do we define constraint so that we can include that in
the defintion
michaelS: All info?
benws: Enough!
... How do we design constraints so that there is enough
info
RESOLUTION: ODRL constraints SHOULD be self contained and not depend on additional prose, noting that it may be a subset of a larger human readable policy.
PROPOSED: A constraint SHOULD carry enough information to check whether it has been satisfied
<michaelS> +1
<Serena> *I agree with Simon*
benws: The previous proposal is only about whether the solution relying only on the semantics on ODRL. Now we're saying that constraints should be able to check whether they've been satisfied
<renato> About "this long"
<michaelS> Enough = enabling the processor to match left and right operand
renato: I suppose we're saying that in a constraint, we need to be able to refer to a source of info to check whether it's been satisfied. So if the count has to be 10, you find out whether it is or not by looking at {} this URL
simonstey: Maybe then we say that a constraint should provide info for checking that it has been satisfied
benws: Can you have a constraint like 'read the licence'
simonstey: That's why it's SHOULD not MUST
<michaelS> "enabling to match left and right operand" is enough
PROPOSED: A constraint SHOULD
carry "enabling to match left and right operand information to
check whether it has been satisfied
... A constraint SHOULD carry information to enable the left
and right operands to be processed to check whether it has been
satisfied
<victor> +1
michaelS: I've proposed to use
the term matching left and right operand by the rule of the
operatr
... I agree to the proposal, but I understand the concern about
processing.
<simonstey> +0.5
<CarolineB> +1
<Serena> +1
<Sabrina> +1
<michaelS> +1
phila: I'm happy with this as a WG reolution, but we need to be careful in the spec itself
<ivan> +1
<renato> +1
<victor> +1
RESOLUTION: A constraint SHOULD carry information to enable the left and right operands to be processed to check whether it has been satisfied
benws: I think we're in agreement
that we have a real requirement, but we may not have defined
how to solve it.
... We have a req for complex constraints and these 2
resolutions go some way to defining what the resolution
[Coffee Break]
<Serena> *I'll have to leave at 13:30 CET for seeing the doctor, so I'll leave my scribing duties a bit earlier to Simon. I hope to be back asap then.*
<Serena> scribe: Serena
renato: try to summarize
constraints on constraints
... we agreed to have multiple constraints on constraints
<renato> odrl:constraint [
<renato> a odrl:Constraint ;
<renato> odrl:subject odrl:play ;
<renato> odrl:deliveryChannel <https://iata.org/aircraft> ;
<renato> odrl:operator odrl:eq
<renato> odrl:source <http://my.co/getTheDeliveryChannel?now>
<renato> ] ;
renato: here there is a very
simple example
... two new predicates
... odrl:subject and odrl:source
... enough information to process left and right operands
benws: example of constraints on constraints using subject?
<simonstey> +q
renato: it is useful also with multiple actions in a permission
michaelS: looking at subject, I'm not sure it is really of help
<phila> scribeNick: Serena
ivan: I'm a little bit unsure
about the meaning of subject vs. source
... when I talk aboyt constraints in general the subject is
what I constrain
... while here it seems to be the source
renato: no you constrain the "play"
ivan: I create a constraint on an
instance
... here the subject is a class?
... it's not an individual
... I'm looking for a very clear statement
<renato> odrl:action odrl:play ;
ivan: I'm not sure to understand
if I read it from ouside
... we can create a constraint with everything which has an
id
... which one is the id of the instance here?
... it is probably source right?
renato: odrl:play is a predicate defined for the action
<renato> odrl:permission [
<renato> a odrl:Permission ;
<renato> odrl:assigner <http://socan.ca/> ;
<renato> odrl:target <http://socan.ca/catalogue> ;
renato: here there is the
pre-text
... the asset is the catalogue
ivan: the whole idea is that we have to make it explicit what is constrained
renato: something like "applied to"
ivan: yes
<michaelS> q
renato: source, we will find another better name,
michaelS: I'm confused now, we
said before that we should try to avoid to have all these
constraints
... is it required to have an explicit reference to the action?
Or is it sufficient to say that it refers to the definition of
the constraint?
... is it really relevant what the action is?
renato: right…
https://github.com/w3c/poe/issues/60
benws: can we see this example?
<renato> a odrl:Constraint ;
<renato> odrl:event <the-football-game>;
<renato> odrl:operator odrl:eq ;
<renato> ] ;
<renato> a odrl:Constraint ;
<renato> odrl:appliesTo odrl:event ;
<renato> odrl:dateTime "P30M" ;
<renato> odrl:operator odrl:eq
<renato> ] ;
renato: the appllesTo refers to
the above constraint
... unless we name the constraints
... we name the first constraint c1
ivan: it would be good to see the whole graph
renato: c2 appliesTo c1
ivan: what is the specification
of the domain of appliesTo?
... in c2 you refer to odrl:constraint that is an instance of a
class
phila: I'm worried by object being a triple
ivan: you need to have a name to the permission
phila: c2 shouldn't be listed in
the permission
... because it constrain c1, not pro.com
... there shouldn't be a link from the top block to c2 in my
view
renato: maybe c1 shouldn't be
there
... in the top block
benws: it maybe simpler to have the basic implementation of constraints on constraints
michaelS: c2 does not really applies to c1 but to the aircraft of c1
*I'm leaving in 2 minutes, simonstey could you please take my place now?*
<simonstey> scribe: simonstey
<Serena> *htnaks simonstey I'll be back asap*
<Serena> *thanks*
[discussing about constraint scoping]
<phila> scribeNick: simonstey
renato: [switching from airplane example to datetime/event example]
michaelS: for me it doesn't really make sense.. how do those constraints relate to each other?
ivan: we are dangerously close to have something requiring some sort of event model
phila: well it's something we need to solve somehow.. whether that's the right solution or not
+q
-q
benws: maybe we should have a property called embargotime?
+q
benws: I think it is a good example.. eventually it may be pushed into a profile, but for now I think we shouldn't be afraid of redefining semantics
{michael & renato talking about possible solution approaches]
<phila> simonstey: I'm wondering where we draw the line between generic and specific use cases
<phila> ... I don't think it makes sense to come up with a property that refers to the end of an event specifically
<phila> benws: If we allow ourselves to be free on the semantics for now, we see how we end up after working on an example or two. Maybe start with pay amount example
<phila> renato: So you want to get rid of A$ as a constraint
renato: what if the unit is a constraint term itself?
ivan: relating to something we
discussed earlier (the football example)
... we don't know what an event is
... ODRL doesn't define what an event is
<ivan> http://motools.sourceforge.net/event/event.html
ivan: e.g., for events we may
define properties that we can constrain
... maybe taking the event ontology as guidance
benws: can we take the unit out and make it a seperate constraint?
<phila> ]ack me
<Zakim> phila, you wanted to talk about blank nodes and to talk about odrl:payAmount [rdf:value 100; odrl:unit "USD"]
phila: the value of payamount is defined as xsd:decimal
<phila> odrl:payAmount [rdf:value 100; odrl:unit "USD"]
phila: for rdf that means we could use a blank node
benws: but this would require
defining complex objects for each and every situation
beforehand
... but this would require defining complex objects
... for each and every situation beforehand
-q ...
renato: we also have to think
about the general info model
... we don't only have RDF but other serializations too
... [working on webex whiteboard example]
ivan: how do I properly define
the meaning of odrl:unit in that case?
... what's the definition of currency?
... it doesnt really define what the currency/unit constraint
really mean?
... how do those two constraints exactly relate to each
other?
... how do I determine the leftoperand of :c5
...: c5's odrl:eq refers to a property that isn't contained in
:c5
... there is no clear definition on what a constraint applies
on
... applying it on a class instance as we currently do, doesn't
cut it
michaelS: I feel that what you talked about ivan relates to what we did already, i.e. going through all current constraint definitions and refurbish/clarify their descriptions
benws: we want the rightoperand of c4 to say 50aud(?)
phila: i think that's actually
not a constraint itself but more of an extension
... to an existing constraint
<victor> +Q
victor: 1) we have an ontology,
so we have the opportunity to be precise
... 2) we have agreed before the break that we won't be able to
have complex constraints on complex constraints
... iirc
... but haven't we said we wont have a constraint on a
constraint on a constraint .. ?
+1
<victor> -q
phila: the thing that worries me
about ODRL is trying to enumerate possible solutions
... which would be outdated the moment you are publishing
it
ivan: you can define your own
datatypes..
... which might be dependend on particular profiles
phila: I would be happier to have
complex objects instead of all the things we try to
enumerate
... and/or having constraints on constraints on
contraints..
<victor> +1
renato: where does that leave us?
ivan: maybe pushing the
complexity somewhere else
... i.e. keeping the current 1 layer constraint model
michaelS: some leftoperands are covering spatial constraints
+q
phila: there are defined regions/subregions but you won't need geometry
renato: I'll try to summarize
<renato> https://www.w3.org/2016/poe/wiki/Constraints
renato: seems as if we will try
to keep the status quo
... improve/extend constraint concepts as described in the
respective wiki page
... does that mean we would need to deprecate e.g.
odrl:unit?
... I've to check how that translates to XML
... github issues 62 & 60 will be updated, i.e. that we
improve the definition of constraints
<victor> +1
benws: aren't they two different issues?
renato: 62 is constraints on
constraints
... 60 is about the subject of a constraint
<renato> https://www.w3.org/2016/poe/wiki/Requirements#POE.R.DM.02_Define_target_of_a_constraint
renato: trying to figure out what this requires us to change
<renato> https://github.com/w3c/poe/issues/59
renato: constraints may apply to
assets/parties too
... so it's probably more related to issue 59
ivan: I think it's better to have
everything we decide/agree too to have also persisted on
github
... maybe change the label to editorial
... but first we have to agree here
<renato> Proposal: For Issue#60 we work on improved Constraint Definitions and include non-simple values for some constraints where it makes sense
<ivan> +1
<phila> +1
<CarolineB> +1
<Sabrina> +1
+0.5
<James_> +1
<renato> +1
<michaelS> +1
<victor> +1
RESOLUTION: For Issue#60 we work on improved Constraint Definitions and include non-simple values for some constraints where it makes sense
<michaelS> back at xx:40
<Sabrina> yip
renato: we continue with issue59
<Sabrina> im eating lunch sorry
<scribe> scribe: victor
<scribe> scribe: victor
<simonstey> scribeNick: victor
<renato> <http://socan.ca/tariff:13A>
<renato> a odrl:Offer ;
<renato> odrl:permission [
<renato> a odrl:Permission ;
<renato> odrl:assigner <http://socan.ca/> ;
<renato> odrl:target <http://socan.ca/catalogue> ;
<renato> odrl:constraint [
<renato> a odrl:Constraint ;
<renato> odrl:subject <http://socan.ca/catalogue> ;
<renato> odrl:language <urn:ietf:bcp47:fr> ;
<renato> odrl:operator odrl:eq
<renato> ] ;
<renato> odrl:action odrl:play ;
<simonstey> +q
<phila> ack m]
michaelS: in the ODRL community,
it was said "this will be defined by the definition of the
assignees"
... for example, adults being +18 would be part of a specific
profile, as it makes the process quite complex.
simonstey: we have enough
constraints concepts that can be checked.
... constraints for actions mean "if the constraint is not
valid, the permission is not active".
... whereas we want to remove certain assets from a group of
assets etc. Shall this remain up to the implementors? Shall we
define it?
renato: the current model for
parties at least has the concept of "scope" (attribute that
indicate the context)
... having for the moment at least the values "individual" and
"group". but nothing prevents from having other URIs, like
"http://has18year.old"
benws: so, again, the question is whether the complexity is within or outside ODRL
renato: but scope does exist, at least for parties --not for assets though.
simonstey: this adds extra
complexity for formally defining the semantics
... possible mess with individuals/groups and policies
differently declared for them.
renato: the group scope for the
party is used so that "each individual member of a group"
should see the same policies, the policy applies for each of
the members.
... so the problem was already there. maybe we can add scope to
the asset
... Two alternatives: constraint model vs scope model
<Sabrina> +1 for scope
<michaelS> +1 for scope
<benws> +1
<CarolineB> +1 scope
<James_> happy to explore
renato: proposal to augment the scope attribute to asset as well
michaelS: scope is more about how to interprete the group.
victor: my concern is about inconsistencies
+1
victor: we should have a list of
"actions" to be made on policies, whose behaviour should be
defined and tested in the testbeds.
... like for example "inconsistency detection".
RESOLUTION: to add the scope mechanism to the assets.
renato: does anybody object
to
... unit of count expressed as a new constraint?
PROPOSED: to create a new righthand operator for unit of count.
RESOLUTION: to create a new left hand operand for unit of count.
<benws> +1
<renato> +1
+1
<Sabrina> +1
<michaelS> +1 - for a left hand operator
<ivan> +1
<simonstey> 0
<CarolineB> +1
Issue56: rightOperand is a reference to a value
https://github.com/w3c/poe/issues/56
<renato> https://github.com/w3c/poe/issues/56
renato: shall we overload the
meaning of datatypes like xsd:anyURI?
... we may say that if we find this datatype we should
understand "go there and retrieve that".
xsd: string
... anyURI
victor: in RDF I will be using URIs very often.
renato: we can distinguish both cases with datatypes.
sabrina: why dont we use webId?
<Sabrina> https://www.w3.org/wiki/WebID
phila, do you mean https://www.w3.org/2001/tag/group/track/issues/14?
ivan: webid is not widely used
sabrina: ok. agreed.
ivan: concerned but in a imprecise manner
sabrina: in lisboa we were always thinking about the support to XML
renato: but after we polled, we agreed that XML was in use
phila: mostly agrees with ivan,
but also in an unprecise manner
... i prefer calling it "API call"
renato: what is the working group feelings?
phila: there might be problems if
a dumb parser finds the URL and does not know that it is an API
call.
... perhaps this can be done with a new property. instead of
rightOperand, having rightOperandRef etc.
victor: we may use it in other places
benws: like a moving target
renato: having a "moving action" is not a thing that will happen .
victor: i cannot think that fast, cannot judge
PROPOSED: In order to support API calls, we add a new property rightOperandReference, and a new class
+1
<CarolineB> +1
<michaelS> +1
<simonstey> -0
<Sabrina> 0
RESOLUTION: In order to support API calls, we add a new property rightOperandReference, and a new class
topics: issue 63
<renato> https://github.com/w3c/poe/issues/63
<renato> Earlier resolution (from 17 Oct):: That the extended relations (AND, OR, XOR) will only apply to Duties and Constraints, not Permissions and Prohibitions
renato: We had decided to apply exnteded relations exclusively to duties and constraints.
<renato> https://www.w3.org/2016/10/17-poe-minutes
renato: do you have any comment or questions on that?
<simonstey> c1 xor c2
victor: i support it
<simonstey> c1 appliesTo c2
renato: we are expecting input here.
simonstey: we may have chains of constraints
ivan: i agree that we may have cornercases, but we leave it as a mere possibility
renato: i posed an example on the use of XOR
<simonstey> +q
benws: we have also good examples. although they can be separed into two different constraints
simonstey: it is also relevant for remedies. in the Austrian law, if you do something (steal a car) you can choose the remedy (jail or pay X)
benws: +1
victor: to send a reference to the similar solution adopted in MPEG-21 Media Contract Ontology
RESOLUTION: changes still to be proposed in the information model.
<michaelS> scribe: michaelS
<renato> https://github.com/w3c/poe/issues/48
benws: could we use the version terms from another vocabulary?
renato: listed some examples: draft, approved, ...
ivan: we should look at terms from other standards first
benws: validfrom, validto, isdraft
simonstey: dont' valid... overlap
with constraints?
... a policy as such has no rules =
permission/prohibition/duty
benws: a new policy could retire an old policy
simonstey: not happy about the terms "valid ..."
renato: we can express temporal constraints for permissions ...
<Zakim> phila, you wanted to point to PAV
<phila> PAV
phila: governance and versioning covers what was raised as need
<simonstey> we also have to take inheritance into account
ivan: we should check first which standards provide what this group needs
renato: do we have the knowledge
of other standards to take this action?
... anybody of this groups wants to list the requirements
benws: committed to do that
<scribe> ACTION: List required versioning properties [recorded in http://www.w3.org/2016/12/05-poe-minutes.html#action01]
<trackbot> Error finding 'List'. You can review and register nicknames at <https://www.w3.org/2016/poe/track/users>.
renato: went over editoral issues at https://github.com/w3c/poe/issues/48
<phila> ACTION: ben to list required versioning properties [recorded in http://www.w3.org/2016/12/05-poe-minutes.html#action02]
renato: correct URL https://github.com/w3c/poe/issues
<trackbot> Created ACTION-38 - List required versioning properties [on Benedict Whittam Smith - due 2016-12-12].
renato: started discussing https://github.com/w3c/poe/issues/80
<simonstey> +q
simonstey: disadvantages are more
relevant than the benefits
... of having multiple actions in a rule
benws: worried about permission explosion
<CarolineB> *me Really sorry - have to go for anappointment
benws: can think about applying 30 permissions by a policy
<simonstey> The Child Policy MUST override the Parent Policy. i.e.: If the same Action appears in the Parent, then it is replaced by the Child version, otherwise the Parent Actions are added to the Child’s Actions.
<simonstey> http://w3c.github.io/poe/model/#inhertiance
renato and benws: if constraints apply only to a subset another permissions has to be created
benws: typically the same constraints apply to all actions permitted form an asset
simonstey: re the previous discussion: doesn't this makes a permission complex
benws: listed examples for actions which could have the same constraint
simonstey: is afraid of drawbacks
not noticed yet
... not sure if having three different permission will be
understood in the same way as having one permission with three
actions
renato: this would be a major change to the Information Model, needs a clear definition
simonstey: proposed not to take a decision on that today - people should have time to consider that change
renato: agreed to that - invited
all to go over issue 80
... said we should switch over to the Vocabulary issues
... see https://github.com/w3c/poe/milestone/2
renato: biggest issue is https://github.com/w3c/poe/issues/64
renato: look at the responses of
the user survey
... the open issue is to go over the current vocab and sort out
into normantive and non-normative terms
... should we update the vocab based on results of this
survey?
<Zakim> phila, you wanted to share experience from DWBP
renato: any comment?
phila: we should check who plans to use term - how widely will it be used in practice?
benws: clarified: Reuters Media
is more about news distribution, Thomson Reuters about
distribution of financial data
... should be merged at some point in time
renato: is the requirement for a CR to have at least 2 implementations?
phila: this is a definition by the group - it could be considered in which different business sectors they should be implemented
renato: ok, let's discuss this later
<Zakim> simonstey, you wanted to ask about the necessity of enumerating actions when there's the possibility to define profiles anyway
renato: backt to sorting out normative/non-normative
<phila> My Tweet showing slide by DBPedia use of ODRL, see top left of pic
simonstey: most of the current
Action terms don't have to be normative by his understanding,
only use and transfer could be made normative
... a lot of the definitions of the Actions need to be
reworked.
renato: if an action term is consideres as of wide used it would help to have a definition
simonstey: if he and others from universtities would be involved in the ODRL Community work other terms would be added, e.g. query.
phila: We are not defining an ODRL process, but a vocabulary
<phila> Conformance with ORG
phila: will look at the confirmance statement from another vocab (see url)
<simonstey> +q
simonstey: sees only a need for normative relationships between terms but not for normative terms
ivan: another example: Annotation Working Group - 2 documents: data model and the vocabulary docs
<ivan> http://w3c.github.io/web-annotation/model/wd2/#candidate-recommendation-exit-criteria
<ivan> http://w3c.github.io/web-annotation/vocab/wd/#candidate-recommendation-exit-criteria
ivan: the data model document is much more strict, the vocab is only strict about some basic requirements, less about the terms
michaelS: IPTC sorts our data models (normative) and vocabs (non-normative), but some vocabs are essential for the data model - like e.g. the Policy Types for ODRL
ivan: we should not sort out individual ODRL terms but terms for a specific - required - purpose
benws: liked this principle - think about the role of Profiles in ODRL
renato: look at the survey results and terms and keep only the most wanted terms
<simonstey> +q
phila: do we have a cutoff-date for that?
renato: we have a timeline agenda items
<renato> http://w3c.github.io/poe/model/#profile
ivan: fine with that approach. We should add a non-normative text about how profiles should be created and look like
renato: shared the link for that
<simonstey> +q
renato: we are in a starting phase about profiles
ivan: We need an explicit statement that this document is not normative
simonstey: we should explain how to add their own action terms to a basic ODRL Action Vocab
michaelS: we have SKOS
simonstey: we may need more than the SKOS relationships
<Zakim> phila, you wanted to talk about profiles
phila: re profiles: run a workshop about vocabulary profiles ... will discuss many facets of that, a working group for that should be established in 2017
<renato> http://w3c.github.io/poe/ucr/
simonstey: has intensively worked
on the Editors Draft http://w3c.github.io/poe/ucr/
... this document is in line with what we have on the Wiki page
https://www.w3.org/2016/poe/wiki/Requirements
renato: this document looks good
benws: asked to set a clear due date for all documents
renato: so far no real activities on the Formal Semantics and the Best Practices documents
UNKNOWN_SPEAKER: the question is: what will happen after having a CR?
phila: before going to CR: approve people outside have looked at it. The group should try to approach people to get wide reviews - and this should be documented.
ivan: let's use Github for that
phila: the group should respond
to the review of a reviewer - and the reviewer should be happy
with the comment
... finally no standing issues are allowed
... The W3C director may be interested in having many reviews
and less in the content
ivan: added: we have to contact
the horizontal reviewers in other working groups soon
... as early as possible
... security issue: the RC document must include security
considerations - we have to check what's a security issue
... every week counts
renato: should we create a list of about 50 persons and invite them?
phila: we could create a template email for this purpose
renato: how long takes a horizontal review?
ivan: unpredictable - it may take a long time. We should tell them when POE wants to go to a CR - suggested to plan 3 months for that.
renato: is the planned f2f in March to early?
ivan: inclined to say yes.
phila: consider Easter in
April
... proposed the week of 24 April
ivan: feels that a CR could get
ready in mid-May
... in summer time implementation reports shoudl not be
expected
... PR could be expected in October
... the group needs to define a testing scenario and share that
with the implementers
... have a look at what the Annotation WG - this group may
reuse some parts of that set up.
... we need a person of this WG who commits to be behind all
the testing and implementations
renato: thanks for this imput. As
a next step we need a timeline with all the required
steps.
... this will result in much more work by this WG as
expected.
ivan: Members of this group should consider who could do testing in their company/organisationb
<phila> phila: Thanks to everyone
<Sabrina> thanks folks....
renato: thanks to all for the input and contributions to this call