add a word about dcat odrl policies#6
Conversation
| odrl:assigner <http://ds.decide.lblod.info> ; | ||
| ] ; | ||
| odrl:prohibition [ | ||
| odrl:action odrl:write ; |
| odrl:prohibition [ | ||
| odrl:assigner <http://ds.decide.lblod.info> ; | ||
| odrl:target private-ds-ex:dataset ; | ||
| odrl:action odrl:read, odrl:write ; |
|
|
||
| #### Unifying dataset ODRL policies and mu-authorization ODRL policies | ||
|
|
||
| The authorization policies for mu-authorization expressed in ODRL are currently not connected automatically to the datasets published in the data space. This is not straight forward to enforce for all types of distributions either. The authorization policies make sense on the level of a SPARQL endpoint, but are much to fine grained for describing access to simple download distributions. On the other hand, the ODRL policies related to a dataset may specify information on what can or can't be done with a dataset's content, like republication or even reselling the dataset. These things fall more into the legal domain than in the software system authorization domain. |
There was a problem hiding this comment.
Can you elaborate a bit more on the first sentence? I think we should write out a some examples what a publisher needs to do to align the DCAT policy with the mu-auth policy.
Case 1: I defined a public DCAT + everyone can read: what does this look like in mu-auth
Case 2: I defined the licensed DCAT + only group X can read and modify: what does it look like in mu-auth?
There was a problem hiding this comment.
explained why it is not possible to do what you ask in the text.
| The authorization policies for mu-authorization expressed in ODRL are currently not connected automatically to the datasets published in the data space. This is not straight-forward to enforce for all types of distributions either. The policies defined in mu-authorization make sense on the level of a SPARQL endpoint, but are much too fine grained for describing access to simple download distributions as they restrict reading and writing on specific graphs and even types and predicates in those graphs. On the other hand, the ODRL policies related to a dataset may specify information on what can or can't be done with a dataset's content, like republication or even reselling the dataset. These things fall more into the legal domain than in the software system authorization domain. | ||
|
|
||
| Authorization ODRL policies are enforced by the mu-authorization component, this is not easy for the ODRL policies describing datasets in a DCAT catalog. Given the federated nature of the data space, some distributions may not be under control of the system but rather reside in the system of a participant, fully outside the scope of the system that holds a DCAT description of the dataset. | ||
| It isn't even easy to set out some general guidelines on what would be required of a mu-authorization ODRL policy for a dataset that is marked as 'public' in the ODRL policy attached to its DCAT description. One can expect there to be only `odrl:read` actions in its policy, and now `odrl:write` policies, but mu-authorization will be *much* more precise in which specific types and predicates to expose in its data slices and this will be quite custom to *which parts exactly* of the dataset will be exposed. The [#final-semantic-components-and-why-if-any](write-up-odrl.md#final-semantic-components-and-why-if-any "mention") has some examples of just how specific these rules can be. Needless to say Mapping a 'private' dataset to a mu-authorization mu-authorization policy allows for even more degrees of freedom. |
There was a problem hiding this comment.
"and now odrl:write policies", what do you mean?
Use odrl:modify
There was a problem hiding this comment.
Fix "dataset to a mu-authorization mu-authorization policy"
There was a problem hiding this comment.
typo: changed to and no odrl:modify policies
No description provided.