Re: [ogpx] OGPX Draft Charter, 2009 08 25 edition

[Infinity Linden <[e-mail]>]

oh, one more point about IM. lemme cut out everything else but your
relevant comments on IM.

On Thu, Aug 27, 2009 at 4:02 AM, Carlo Wood<[e-mail]> wrote:

> Correct me if I'm wrong, but from this I conclude that your vision of
> a virtual world is a collection of regions that all use a single point
> (and administration) for authentication of a users avatar (and inventory).
>
> For example, one logs in with the agent domain run by LL and than can
> move between regions and IM every other avatar in "the virtual world"
> (using the handle of this agent domain provider). If one would prefer
> to login with a different agent domain provider, it would not be possible
> to IM every other avatar, or have other interactions  such as 'giving'
> inventory objects to other avatars, and thus one would not be able to
> be part of the same virtual world.
>

the protocol does not describe the policy for where IM messages may
flow, but rather the mechanism for how the client establishes an IM or
voice connection. it does not REQUIRE systems to operate in the way
you define. and, in fact, one of the motivations to describing a
globally unique URI for each avatar is to allow user agents and client
agent domains a service endpoint to query about agent specific
features like "how do i IM you?"

i think everyone agrees that the existing, legacy user-to-user and
group IM system for second life is sub-optimal, won't scale to the
bazillions of users we want to enable with this protocol and shouldn't
be used in a new design. many people think we should just use XMPP or
IRC and be done with it. and there are some compelling arguments in
favor of this approach.

the work of this group would be to define the mechanism of
establishing, maintaining and closing chat or voice communication
sessions, and the format for addresses used to identify the source and
destination of messages.

this is a very roundabout way of saying, it SHOULD be possible that a
group hosted on a Linden Lab agent domain may contain members that are
themselves registered by a foreign agent domain (let's say OSGrid.)
but before the Linden Lab agent domain would allow messages to flow to
(or from) users with accounts on a foreign agent domain, policy may
require that foreign agent domain to agree to enforce various bits of
Linden's terms of service.

so in other words, i think the prevailing thinking is that we define
mechanisms that defines how to get a message across the wire and how
to identify endpoints so you can use someone's avatar descriptor or
group descriptor to resolve the IP address of the system you want to
start speaking this protocol to. once that is defined, individual
operators will decide what they want to do with regards to letting
messages from foreign agent domains into their systems.

so, in other words, the "virtual world" is also not defined by the
agent domain. it's another reason i kind of dislike the term from a
protocol perspective. it is evocative, but clearly non-normative.

-cheers
-meadhbh