Permissions and Obligations Expression Working Group Teleconference

<Serena> *I'm in a meeting, I cannot join on WebEx for the moment*

<benws> James - are you joining the webex?

<james> trying to join webex, but having network issues

<phila> agenda: https://www.w3.org/2016/poe/wiki/Meetings:Telecon20161212

<phila> chair: Ben

<phila> scribe: Sabrina

<phila> scribeNick: Sabrina

<Brian_Ulicny> +1

Virtual F2F minutes approval

<phila> https://www.w3.org/2016/12/05-poe-minutes

Can we approved the meeting minutes from the virtual F2F?

<phila> Seem fine to me

<Serena> +1

<michaelS> +1

+1

RESOLUTION: Minutes of https://www.w3.org/2016/12/05-poe-minutes approved

Minutes approved

ODRL Information Model

benws: Moving to the ODRL Information Model

Ben lost connectivity... waiting for him to call back in

Ben back

<phila> Issue 61 https://github.com/w3c/poe/issues/61

Inverse target link - if you have lots of assets you are much more likely to point from the asset to the policy rather than visa versa

Any comments?

michaelS: We did exactly something like this in the News Standard. So the question is should this be a recommendation or rather just best practice?
... Shouldn't this be more advice rather than a recommendation.
... The question is who defines this?

benws: We are looking for the inverse of the predicate

michaelS: Where does this reside?

<simonstey> +q

<phila> phila: I'd say that dcterms:license is the way to link an asset to a Policy

phila: I see what Michael is saying... I can see why people would use dcterms:license . It seems like the right thing to do.

<phila> ... It's a commonly used predicate, see, for example, DCAT

<Serena> +1 for dcterms:license

simonstey: This is certainly not the inverse of ODRL:target
... It is very often used in blank nodes where you don't have a permission i.e. rule blank node.
... Unless we move the target from the rules to the policy I would not recommend an inverse relationship

<smyles> http://udfr.org/docs/onto/dct_license.html

<renato> (sorry - just arrived!)

phila: You never know what type of document you get from the predicate. You can only find out from the MIME TYPE

We might want to define text/ODRL

The predicate can't tell you what you get back....

phila: The URI has no semantics....The MIME TYPE tells you what the serialisation is...

<simonstey> +q

phila: In summary dcterms:License is ok

<phila> phila: dcterms:License doesn't tell you what kind of doc you're going to get, that's the job of the MIME type.

<phila> phila: If this Wg wants to define its own MIME type, it can

simonstey: It doesn't make sense to point from the asset to a single rule.... All the rules together form the license...

phila: So you would use dcterms:license to point to the policy

renato: If we have an asset with dcterms:license pointing to a policy and that policy already has an asset in it. What do we do? How can we handle this?

benws: I would add them together

<simonstey> <asset> dct:license/(odrl:permission|odrl:prohibition)* ?rules

renato: Maybe we should consider the template policy - everything without the asset in it

Ben back

renato: My concern is if the policy you point to already has an asset, does the target URI referred to by dcterms:license would it overwrite the asset

<simonstey> +q

benws: If you have target and targeted by and an inverse relationship then reasoning would just add them together

renato: Assume that we are talking about a policy?

benws: No keep it simple and just target a rule

renato: The minimum referable entity is a policy, we can't have rules on their own
... The policy by definition has an asset. If there are 3 permissions then we have 3 assets. If we have an asset that refers to the policy what happens to the 3 assets that have already been defined

benws: Is this relevant for an inverse of odrl:target

renato: No, I am talking about something different

<renato> (old CC/ODRL Profile): https://www.w3.org/community/odrl/work/cc/

michaelS: I recall a similar requirement in a creative commons license

smyles: dcterms:license is not a perfect fit, however I agree with Phil people will use it to point to a policy. There are no ids for permissions and restrictions

<simonstey> +q

benws: This requirements is about the ability to point from an asset to a permission. What do we think of this requirement?

smyles: Can you describe an example of how it would be used?

benws: If you have 10,000 assets then it is very inefficient to point from the policy to the assets. It is far more efficient to point from the asset to the permission

renato: Does this mean the permission does not have any reference to the asset in it?

benws: Yes, this would be the case

<simonstey> odrl:Set ?

renato: The way ODRL was modelled was - I have a policy with an identifier and have references to assets and rules.
... You could use a set

benws: No it's not a set that it required. We want to point from the asset to the policy

simonstey: If you want to define a policy that does not have a target you could use an odrl:set
... Is this an implementation issue. If you want to automatically evaluate the policy and you don't have an asset in the policy you would need some guidance as to where to look for the asset
... This might work in an inhouse scenario but not in general

benws: You just do a query

simonstey: But what do you query?
... You don't know what is pointing to the policy and therefore you don't know where to query

<renato> XML encoding rules: http://w3c.github.io/poe/vocab/#xml

smyles: We have the same type of requirement and we lobbied for this is be included in the past. Possible suggestions reference to it over and over again, target optional, or whatever refers to it
... I am in favour of solving this problem

<renato> PoWder !!!

<renato> :-)

phila: This sounds very familiar. One set of metadata to an undefined thing and it is governed by whatever points to it
... In POWDER it was possible to do this limited by domain name (i.e. put an outer wall on it)

renato: In order to solve the problem - given a policy that we want to use as the target for these statements - should we allow the policy not to have an asset in it

<simonstey> +q

renato: from your asset list be it 50 million of them, we don't have any assets in the target policy

benws: They are implicitly there
... If you were to materialise the triples then indeed the policy would have a target

renato: In an implementation yes

simonstey: An agreement policy is not only about the asset which is referred to by the rules, actually as far as I know you can have multiple assets in each rule. The same for assigner and assignees.
... When you have 200 million assets you would have to repeat the assigner and assignee multiple times in each rule

benws: This requirement does not refer to assigner and assignees

simonstey: But you would have to repeat this

benws: Ya we would repeat asigners and assignees in the rules

<simonstey> Must contain at least the Party entity with Assigner role and a Party with Assignee role. The latter being granted the terms of the Agreement from the former.

<simonstey> https://www.w3.org/TR/vocab-odrl/#term-Agreement

renato: When assets that refer to policies where there is no asset in the policy we need to indicate that this was done on purpose
... otherwise people would expect an asset

Brian_Ulicny: Can't we use a class here

<renato> see "Asset: the Permission entity must refer to an Asset (where at least one, and only one, relation value is target) on which the linked Action should be performed (required)"

<renato> In the Information Model for Perms and Prohibs

<phila> Anything that points to this policy is covered by it

smyles: Couldn't we use a url to indicate a class. This is what we did at AP. We don't know the ids of the assets. We put in an identifier the meaning is that anything that points to this identifier is governed by the policy
... The processing engine needs to figure got if this asset is governed by this policy. We need an engine that can evaluate this.

benws: When in doubt add a level of indirection

F2F meeting

benws: As our hour is up, we have to bypass the other items on the agenda

<Brian_Ulicny> Sorry. Have to drop.

<phila> Proposal: F2F meeting 18-19 May 2017 in London

renato: The F2F will now be in May - there are a number of events in London in May - so we propose the 18th and 19th of May in London

<benws> +1

<phila> (instead of Madrid or Vienna)

<phila> +1

<renato> +1

+1

<ivan> =1

<smyles> +1

<simonstey> 0

<ivan> +1

RESOLUTION: F2F meeting 18-19 May 2017 in London

IPTC meeting will be held there....

<phila> [Merry Christmas]