Re: [ogpx] OGPX WG draft charter, 2009-08-19 revision

[Morgaine <[e-mail]>]

Carlo Wood has once again done a magnificent job of presenting the key
issues.  I'll get to this in a minute.

I've stayed away from the discussion for a day while you all battled away at
defining terms, not because I'm not interested in terminology (I'll be as
pedantic as the rest of you once we get to the drafts :P), but because I'm
attempting to focus on the end goal without getting distracted by the means
to achieve it.

The end goal is of course to determine whether the OGP *mechanism* will
handle interoperation between SL-type virtual worlds (independent
collections of regions), or not.  *How* that is achieved, and what
terminology is used to describe it, is of no interest whatsoever compared to
the actual answer, in this context.

I am thrilled that the consensus appears to be moving towards unanimous
agreement that such interop via this protocol is indeed the end goal.

Some have suggested that it was always the end goal of OGP, but the highly
ambiguous and rather eclectic nature of the terminology in the charter
easily exposed the workgroup to eventual rejection of such a goal as being
"out of scope".  The last day's deliberations have been invaluable, because
they have ensured that this end goal is indeed in scope for the workgroup
and hence cannot be dismissed or minimized.  This will have a significant
impact on our work and our deliverables.

Carlo has again performed a masterful analysis of the situation.

I agree with every word, except that my personal impression that "*Linden
Lab wasn't interested in the slightest to have this kind of interoperability
with any of the current opensim pioneers*" refers to interop policy and
hence misses my focus.  It should be replaced by "*Linden Lab did not
specify this kind of interoperabiliy in the OGPX charter and therefore it is
unsafe to assume that this interoperability will be in the OGP protocol"*.
These are quite different things, and my interest is in the latter, the
mechanism of OGP, since I am focussed on technology.

Now that we know that OGP is intended to provide such interoperability (if
there is still disagreement about this in any quarter, please speak up now),
we can make provision to ensure that the work of OGPX does actually deliver
it.

There are numerous areas impacted by the decision to provide such
interoperation, but the charter is not the place to describe them.  However,
the charter must provide food and shelter for this area of work, otherwise
it delivers a de facto rejection of such interop instead.  In other words,
such interop must be a deliverable.

(I will mention some of the issues that may need addressing to achieve this,
not because they are relevant now, but because it demonstrates that the
currently listed deliverables do not necessarily cover them.  Included are
(this is *NOT* exhaustive):  interaction between two or more Agent Domains,
multi-domain identity handling, symmetric relationships between peer worlds,
composition of inventories from multiple agent domains, multi-world
addressing or landmarks for cross-world teleports, multi-world currency
where available, open trust relationships, and transitive trust
relationships.  In particular is must be pointed out that full connectivity
is not viable in an internet-scale metaverse, and this impacts on the core
design.)

Without going into the detail at charter time, I hope it is nevertheless
clear that delivering cross-world interop is not trivial.  In order to
ensure that it does not fall victim to accidental (or intended) disinterest,
it needs a specific deliverable to be listed among the Goals and Milestones
detailing cross-world operation in overview, and it also needs every other
deliverable to consider cross-world issues in passing.  It should be noted
that, if OGP was always intended to support such interop, then *the latter
will entail no extra work*.

As Carlo said, when we state that cross-world interop is the goal, "We're
not there yet".  We also have to provide care and feeding for this, to make
the goal a genuine one.


Morgaine.





===================================

On Sat, Aug 22, 2009 at 3:13 AM, Carlo Wood <[e-mail]> wrote:

> On Thu, Aug 20, 2009 at 05:16:09PM -0700, Meadhbh Siobhan wrote:
> [deleted what I agree with, or just needs no comment]
> > one of the objectives of the OGP protocol is to allow a virtual world
> > to be constructed from regions, agent hosts and asset servers from
> > different administrative domains. in order to do that, one of the
>
> Okaaayy... now you're using "a virtual world" again for a collection
> of regions run by *different* administrations, while I was proposing
> to keep the fact that currently all different (Second Life like)
> worlds are run by different administrations. But... I'm not in a hurry
> to have the meaning of "world" clearly defined, the only thing
> that is important for me is the intent of the people on this mailinglist
> and those working for Linden Lab in particular ;)
>
> > things we think we need to do is to have a scheme by which regions are
> > unambiguously identified with a URL. we don't currently list
> > "unambiguous landmarks" in the charter because we don't talk about
> > landmarks at all in the charter. it's also unclear whether a landmark
> > format will be defined as part of this effort.
>
> Okay. Note that I only talked about landmarks as example with
> as goal to get clarity between us, here and now; followed by
> a search for correct wording of paragraphs. I never thought that
> "landmark" had to be added to the charter in said paragraphs.
>
> > so i guess what i'm saying is that if we listed "a format for
> > unambiguous landmarks" in the charter, shouldn't we also list other
> > Second-Life-like features like the C/M/T permissions system? the
> > unambiguous URL representing a region is required for proper operation
> > of teleport, which IS in scope, but the unambiguous landmark format is
> > not required as landmarks are concepts more appropriately manipulated
> > by client applications.
>
> I don't think that even unambiguous URL needs to be mentioned.
>
> > so, is it enough to add verbiage to the charter indicating that the
> > protocol requires regions to be addressable by unambiguous URLs, but
> > not explicitly constrain their use in the charter?
>
> Again, it was just a down-to-earth example, not suitable for the
> charter.
>
> The problem that we're trying to tackle, that I was trying to tackle,
> is this (going to formulate this VERY explicit, almost embarrashingly so
> ;).
>
> There are people on this list that run (currently) small grids
> with (obviously) opensim servers. They would very much like it
> to be possible for people in SL to teleport directly to their
> grid (and back), preferably (but not necessarily) while keeping
> access to their inventory and keeping there appearance intact.
> The reason they want that is simply: the more seamless that is
> possible, the more people will come; and more people is Good(tm).
>
> The likeliness that this is going to happen for them in the
> future depends mainly on two things:
>
> 1) Will the protocol support it at all
> 2) Will Linden Lab grant those grids whatever access they need
>
> Point 2 is policy, and has nothing to do with the efforts of
> this group.
>
> It is my understanding that Morgaine after reading the charter
> lost all hope to achieve this goal of interoperability, and
> concluded that with the given formulation (as proposed by you)
> *apparently* Linden Lab wasn't interested the slightest to
> have this kind of interoperability with any of the current
> opensim pioneers; which led to some emotions and rather direct
> attempt to let you "confess" that LL wasn't interested, and
> when you said "that doesn't belong in the charter" that was
> interpreted as "we have a hidden agenda".
>
> At this point I believe that all of that has been a misunderstanding,
> that is directly caused by the confusion that exists about what
> "virtual world" means. Morgaine (and me) interpreted "world"
> as "run by a different administration", while you seem use the
> word for "all regions that one can teleport to". Obviously,
> the latter says NOTHING about the intent of Linden Lab to
> interoperate with any opensim grids :p, and I suppose you
> don't have to tell us if they do. But it's good to know now
> that the protocol will support this.
>
> We're not there yet though:
>
> The reason that two grids (or virtual worlds, or different
> administrations) might decide not to allow teleporting is
> probably for a large part going to be determined by (not) trusting.
>
> Therefore, in order to achieve the goal of teleporting between
> Second Life and the smaller opensim grids, we'd like a protocol
> that demands as LITTLE trust as possible.
>
> For example, if the protocol would define something like:
> - in order to teleport, grid A logs into auth services of grid B
> once and then has access the everything after that (including
> allowing to teleport); then teleporting would be impossible unless
> grid B would trust grid A completely.
>
> Other shemes are thinkable that require less trust. One result
> of a "hidden agenda" could be that the protocol would be designed
> such that ANY operability would only be possible with FULL access
> to everything, and therefore full trust, and therefore (in practise)
> no interoperability at all (except between large companies with
> lawyers and contracts-- not the little pioneers that run opensim
> grids now).
>
> --
> Carlo Wood <[e-mail]>
> _______________________________________________
> ogpx mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/ogpx
>