Access Control
Access Control is a mechanism through which an agent ( an HTTP server in this case ) permits other agents --
individuals, organizations, and/or groups made up of these -- to perform certain operations on resources as
specified by policies for the resources and for the agents. Within this document, the resources are LDP resources, but the access
control may operate at different granularities: RDF or other documents, named graphs or individual triples.
The operations are typically create, read, update, and delete (CRUD) but other operations can easily be accomodated by
this design.
When an agent requests a collection of resources it gets to see only those resources or parts of resources
it is authorized for.
Depending on the granularity, the access control mechanisms may affect performance, but should not affect
semantics.
For access control to come into play, the server must restrict some operations on some resources.
Usecases
Access Control on manipulation of resources via HTTP
Adam's user agent attempts:
- To CREATE, READ, UPDATE (or PATCH), or DELETE a resource identified by a URL. The server may immediately
allow or deny the request, or it may request that he authenticate to confirm his privileges,
as specified by the ACG for the Resource.
- If he is denied access, an explanation of why all or part of his request was denied should be provided
so that it becomes possible to detect errors, and so that he may modify the request --
potentially to include making a request for such privileges.
- Adam would ideally like to know whether he will be able to perform an Action on a Resource
before attempting such - i.e., whether he will have to authenticate before he is able to Read or Write the Resource.
Editability of Access Control Rules using HTTP
-
Bart's user agent logs on to a server and requests
the capability to read a group of related resources such as all the papers presented at a conference.
- Employees with job titles VP or SVP can sign (update) supplier contracts.
- Charlie, the Webmaster, would like to grant read access to the papers presented at a conference to all the
people who attended the conference.
User Interface Scenarios
Eddie's HTTP based user agent would like to provide a user interface to allow, where possible, Eddie to
- Know if he can edit or delete a resource.
- Know what he would have to do to have access to a resource ( be someone's friend, be part of a club, have
paid a fee )
- Allow Eddie to edit the access control rules for a resource such as:
- Allow friends of his to access a document.
- Allow friends of his to POST to a container, but only read a subset of the contents of the container,
those posted by that agent for example.
- Allow all the members of the LDP WG to create and edit resources including LDP Containers under a
specific URL pattern.
- Allow all friends of friends as expressed by the foaf:knows relations in one's foaf profile to POST comments
to a container related to some content, and edit their own comments.
- Allow the members of the LDP WG, the RWW CG, the WebID CG, and the member of the European Ontologist Network,
to work together on set of ontologies. It should be possible to drag and drop URLs for these groups,
found on the web, onto the User Interface as a way of creating the union of the members of the group.
Requirements
- An Agent must be able to authenticate itself to a server with an identifier or as the owner of a token.
( All use cases )
- Ability to specify a collection of agents, identifying agents with URIs, URI patterns, or by description.
(Usecase 3.2.2, 3.2.3)
- Ability to specify a collection of resources, identified by URIs or URI patterns or by description,
with a specified access policy. (Usecase 3.2.1, 3.2.3)
- Ability to connect a collection of agents with a collection of resources with given access privileges.
( All use cases )
The above requirements require the ability, by an authorized agent, to CREATE, EDIT, UPDATE relevant ACGs.
- Ability to specify access privileges at a fine-grained level. (Usecase 3.1.2, 3.2.1.2)
- The server should be able to describe access control policies for a resource. (Usecase 3.1.4, 3.3.1, 3.3.2)
- The server should be able explain the reasons for access being disallowed in a machine readable format.
(Usecase 3.1.3)
- A user-agent should be able to find the ACG for a given resource.(Usecase 3.1.1)
- The ability by one user agent to delegate the authority to create and edit ACGs to another agent.(Usecase 3.3.3)
Outline of a Charter for a Access Control WG
An Access Control Graph (ACG) consists of two kinds of collections: a collection of agents and a collection of
resources. It then connects a collection of agents with a collection of resources with the connection identifying
the privileges the agents have on the resources: CREATE, READ, UPDATE, DELETE.
ACGs are resources in their own right and can have access control priviledges specified for them just like
any other resource. This permits the creation and modification of ACGs to be delegated.
The members of the collection of agents contain tokens that the agents obtain from some authentication service.
The members of the collection of resources are URIs or URI templates.
The WG will need to decide whether it also wants to define fine-grained access control at an attribute level.
Deliverables
- Define the collections that are part of the ACG and define how a collection of agents is connected to a
connection of resources.
- Define how ACGs can be created and edited and how these rights can be delegated.
- Describe a proof-of-concept implementation of how a request for access to a resource by an agent can be processed
efficiently with the ACG structure defined above.