Access Control Teleconf 20080509

Notes from 20080509 Teleconf

Attending: Jim, Brian, Mary, Paul, Drummond, Markus, Valery, Bruce, David, Tony, others I missed?

Agenda:

• Review http://wiki.eclipse.org/Access_Control_in_IdAS.
  • Specifically http://wiki.eclipse.org/Access_Control_in_IdAS#Current_Proposal

Notes:

• Looking at Policy-Entities.pdf
  • Paul intended this to be a simple proposal and to show how it might help Parity do something they're working on.
• Tony: Bootstrap issue
  • What's the policy on the policy
    • need to make sure we can make policy statements that refer to policy statements
  • In a new ctx, what are the default policies?
    • Who has rights to create the first policy objects?
• David: Where are the enforcement points?
  • Jim: The CP must ensure (whether it has to do it, or the backing store does it)
  • Paul: Agree
  • David: If there's an external enforcement engine, I have other questions.
  • Paul: There is the option of fronting one CP with another one.
    • In this way, one CP could perform the work of policy management and enforcement. This CP could be stacked in front of another CP that doesn't do this enforcement.
• David: Can the accessing Entity be in another CP?
  • example: cp entities are entries in a database, but the authenticating identity does not exist in the context.
  • Subject ID must come from somewhere
    • CP can derive it from authN materials
      • an authenticated user can be mapped to a subject id
        • In some cases, the mapping is to an entity within the context
        • In other cases, some other subject id will result
• AuthZ Subject IDs need to be flexible
  • need to be able to specify roles, groups
    • example: allow anyone from my facebook friends to do X
  • Drummond suggests using UDI for AuthZ subjectID
• What about referential integrity?
  • The CP must be required to ensure authZ subjectID is not re-assignable
    • In some cases, (i.e. when the authZID is provided by another ctx), it's up to the other ctx.
• No contention regarding the notion of using an access control policy entity
• Next meeting topics:
  • How to perform interrogations like "can bob edit alice's phone number?"
  • David has some others surrounding the policy object.
  • Will we be capable of making negative (deny) statements?