Re: [ogpx] Protocol for permitting policy decisions

[Carlo Wood <[e-mail]>]

On Wed, Oct 07, 2009 at 03:30:26PM -0700, Meadhbh Hamrick wrote:
> a service. but other than to acknowledge that it might happen, i don't
> think it's in our best interest to define a protocol for transporting
> policy descriptions between domains. what policies are possible will
> differ between implementations, and the purpose of defining them as
> part of the policy domain is to explicitly remove them from
> negotiation during protocol exchanges.

I do not agree with this... for some policies this is good, but for
others it is not. Since we won't specify an exhausted list, but -hopefully-
treat everything abstractly, there still is the need for some abstract
list of policy negotiation in the protocol.

Example.

So far we concentrated on the decision "can this user teleport that region",
so allow me to use that case again:

* Problem: user can teleport from wherever to Region1 (R1) if both R1 and
  it's current Agent Service (A1) allow it.
* The Agent Service will initiate this, if it (already) thinks it's ok
  for the user to go to R1 (or would never ask R1).
  So, A1 policies ARE removed from the protocol, as you said.
* What remains are the policies of the region.
  If you eliminate those too from the protocol than the policy decision
  making becomes very unflexible. In effect, the right of the region
  to make also policy decisions would be at the mercy of the agent domain
  operators (ie, responsiveness) and very hard to change, certainly not
  frequently and not at all real-time.
  Surely some policies, like using md5 hashes for something, would be
  fine under such circumstances; but others, like FOR EXAMPLE (we don't
  KNOW what all else might pop up) banning certain users; or toggling
  'region has mature content' (some region might have adult shows between
  20:00 and 4:00, and PG-13 games between 12:00 and 17:00).

-- 
Carlo Wood <[e-mail]>