Re: [mmox] The Story So Far...

["Hurliman, John" <[e-mail]>]

>-----Original Message-----
>From: [e-mail] [mailto:[e-mail]] On Behalf Of Suzy
>Deffeyes
>Sent: Wednesday, February 18, 2009 1:56 PM
>To: Hurliman, John
>Cc: [e-mail]
>Subject: Re: [mmox] The Story So Far...
>
>
>
>On Wed, Feb 18, 2009 at 1:41 PM, Hurliman, John
><[e-mail]> wrote:
>
>
>       The HyperGrid implementation in OpenSim today establishes a full
>trust link between two separate service stacks (grids). It's a very
>clever way of combining grids, but doesn't address how untrusted
>administrative domains interact with each other at all. I don't
>understand why you keep comparing Agent Domain and HyperGrid.
>
>       My understanding of the Agent Domain model is that it maps every
>trust relationship between grids to a policy decision. This is important
>for the business case of company X and company Y creating a partnership
>that is explicitly configured, but glazes over the 99.999% use case of
>moving from untrusted grid A to untrusted grid B. I haven't seen any
>work from AWG addressing the possibility of eliminating all grid<->grid
>communication (google.com doesn't communicate with the links I click,
>the browser is perfectly capable of managing its own state).
>
>
>For the google 2d browsing analogy, I don't need to see a representation
>of other people that also happen to be visiting that same link, and I
>also don't have a name and 3D representation of myself that I want to
>remain consistent on my screen as I go from link to link.
>
>In our MMOX space, it can be desirable to have a consistent
>representation of self that also appears the same to others in the same
>region. Not always, but for a fair number of use cases that involve
>teleportation it does. I'm not sure how that can happen if we eliminate
>all grid to grid communication.
>

"same region" and grid<->grid are separate concepts. The idea of sharing a scene graph across administrative domains is interesting, but the current agent domain work does not include anything like that. Instead, a teleport is done from grid A to grid B where identity and appearance are the only things that need to persist. I don't see why grid<->grid communication is fundamentally necessary to transfer identity and appearance.

>Maybe Agent Domain type implementations 'squirt' too much state along
>with an agent that is being transferred into a region. Maybe we need to
>tease that apart.  I think the actual OGP implementation right now just
>squirts your name, UUID, a few bits, and desired location into the
>region.
>
>Suzy Deffeyes
>IBM
>

>