[Opensim-dev] Global identifiers

Justin Clark-Casey jjustincc at googlemail.com
Wed Sep 1 02:14:09 UTC 2010 

On 31/08/10 18:56, diva at metaverseink.com wrote:
> One last thing, possibly of interest to the folks in VWRAP:
>
> Another thing that happened to OpenSim as it reached 0.7 was the
> reconceptualization of the software itself. As a consequence of this
> reconceptualization, interoperability architectures such as the
> Hypergrid are built as optional components that don't affect the core of
> the walled-garden code.
>
> As such, for those who don't like the Hypergrid for whatever reason, you
> can experiment with your own ideas of interoperability without having to
> tip toe around it. I suggest you understand HG1.5 first, as not to
> reimplement it again. Then, if what you want to do is really different,
> and not just a layer of *policy* above HG1.5 (policy specifications will
> be supported soon), follow the same philosophy of modularizing things
> properly, as that will give your architecture a chance for people to use
> as plugin. If you run into a missing hook, just let me know.
>
> Bottom line wrt interop, two major things happened: 1) the hypergrid
> 1.5, a system-to-system (S2S) architecture with the trust/security model
> explained below; and 2) the clean up of the framework, allowing anyone
> to experiment with interop.

Yeah, I'd really like to +1 this.  If people want to experiment with alternative solutions to Hypergrid on top of 
OpenSim we're very happy to make or accept patches for the required modularity.  And if a core OpenSim developer 
supports and maintains an alternative solution then there's no reason that it couldn't be part of the OpenSim itself, 
provided that it's sufficiently general.  OGP code for instance, is still present in OpenSim.

Hypergrid is a very interesting architecture but I think that Diva would agree that it isn't regarded as the 'one true 
OpenSim way'.  There's clearly a lot of experimentation that has to be done in this whole area before a good solution or 
solutions will emerge (presuming that they are possible at all).

Like Diva, I also think that good standards very often only come out of working implementations.  Hence, though I've 
been following the VWRAP lists (and OGP before that) I haven't been participating since there's been a lot of 
hard-to-follow discussion without much real-world consequence.  And as a working developer I don't have the luxury of 
sitting on my tush and contemplating the Platonic world of future standards all day ;) (joking).

>
> diva at metaverseink.com wrote:
>> One more thing, a bit less important than the others, as the others
>> pertain to grid-level content, and this one pertains to user-level
>> content:
>>
>> - WRT to the user agent itself (i.e. name, appearance, etc.), the
>> user's user agent service (a grid-level service) is the party
>> responsible for creating user agents that are launched at foreign
>> grids. As such, that component is the authority that defines what
>> agent data to send. If the user agent service of one grid so wishes,
>> all of its users' agents can be anonymized and stripped off their
>> clothes before going out.
>>
>> - HG 1.5 has another, symmetric, grid-level component called the
>> Gatekeeper which has the role of deciding what comes in to its grid.
>> So if the Gatekeeper so wishes, it can anonymize all foreign user
>> agents and strip them off their clothes before allowing them in.
>>
>> In other words, the user agent service and the gatekeeper service are
>> the yin and yang of the Hypergrid.
>>
>> diva at metaverseink.com wrote:
>>> [unrelated to the narrow issue at hand, but since people want to
>>> know, here goes]
>>>
>>> HG 1.5 has a trust/security model. The base case is one where grids
>>> are peers, and the traveling of one user agent from his home grid to
>>> another establishes the *base trust* in the following manner:
>>>
>>> - Everything that the agent references from his home grid is made
>>> available to the foreign grid where the user is. In other words, the
>>> user is the driver of trust.
>>>
>>> - Everything else that is not referenced by the visiting agent is out
>>> of reach. "Out of reach" is a soft security model, i.e. the resources
>>> are available on the internet, but you need to know their identifiers
>>> in order to get them. Their identifiers behave as capabilities. This
>>> is the part that still needs work, as Melanie thinks 'soft' is not
>>> enough.
>>>
>>> This trust model is the base upon which trust policies can be
>>> defined. In other words, we now have the basis to add additional
>>> grid-level specifications that overwrite the user's actions.
>>>
>>> Melanie wrote:
>>>> Hi,
>>>>
>>>> HG 1.5 doesn't address these concerns. Also, please remember that
>>>> assets need to be freely available to all, else they can't be
>>>> displayed. The observer gets a copy, too.
>>>>
>>>> Animations, textures, sounds, etc. need to be given to all observers.
>>>>
>>>> Melanie
>>>>
>>>> Mike Dickson wrote:
>>>>> Right. I think some of the use cases related to how content is shared
>>>>> have been glossed over. In a completely open model which is what has
>>>>> been discussed this is all pretty straightforward. But if I'm running
>>>>> an asset service (as part of a grid or separately) I might want to
>>>>> provide access controls as part of that service. The same with user
>>>>> services. I may have a trust relationship with one agent service and
>>>>> allow content to be transferred to agents that service represents. But
>>>>> for another agent service for which no such relationship exists I'd
>>>>> like
>>>>> to deny access to content. And even in the transfer case does the new
>>>>> user get a new copy or a reference. That concept isn't supported
>>>>> now but
>>>>> in a distributed grid its an important distinction. I might wish to
>>>>> know
>>>>> that copies of objects rezzed in a simulator always come from a
>>>>> specific
>>>>> asset service.
>>>>>
>>>>> In short I think how the security model works is way more important
>>>>> than
>>>>> a caching optimization being applied to a URI/URL. Its important to
>>>>> understand what levels of trust between services are supported and
>>>>> under
>>>>> what conditions an access is supported. As an Agent Service I may
>>>>> consider even the "Names" of my users to be confidential and only
>>>>> to be
>>>>> revealed to services for which a trust relationship exists.
>>>>>
>>>>> Mike
>>>>>
>>>>> On Tue, 2010-08-31 at 13:23 +0000, mysticaldemina at xrgrid.com wrote:
>>>>>> Hi,
>>>>>>
>>>>>> As a content creator this concerns me. I believe if I license my
>>>>>> content to
>>>>>> an avatar, and then they go to another grid that any content
>>>>>> pulled should
>>>>>> be from the grid that I have the content loaded into. I think I
>>>>>> should be
>>>>>> in control of my content. I also think I should be able to block
>>>>>> grids that
>>>>>> my content is being accessed from. If you don't always maintain the
>>>>>> original content location there will be no control. If I give
>>>>>> someone a
>>>>>> copy of my content, then that is something else, they are now the
>>>>>> owner of
>>>>>> it and are free to do as they please with it, at least within any
>>>>>> license I
>>>>>> give them. But that is a legal stuff not a technically programmed
>>>>>> one. At
>>>>>> least I don't expect all situations to be programmed.
>>>>>>
>>>>>> Also when asset services start happening this will become more of
>>>>>> an issue.
>>>>>> I will have XRMarketplace.com live soon and plan to start selling
>>>>>> content
>>>>>> and provide that content as an asset server. How will I maintain
>>>>>> any kind
>>>>>> of control over the use of my content if people don't have to pull
>>>>>> copies
>>>>>> from me?
>>>>>>
>>>>>> I also think, and haven't seen in the new hypergrid, if someone
>>>>>> goes to a
>>>>>> new grid I may not allow any of my content to go there unless that
>>>>>> avatars
>>>>>> gets an authorization from me which should be attached to his
>>>>>> proxy profile
>>>>>> for access into my grid/asset server.
>>>>>>
>>>>>> The other thing to think about is how updates or corrections are
>>>>>> propagated.
>>>>>> SL has a terrible system of only supporting copies so any updates
>>>>>> or copies
>>>>>> have to be sent to everyone. Seems content replacement needs to be
>>>>>> supported and if content is all over the place this will get even
>>>>>> crazier.
>>>>>> Also to support dynamic content there needs to be a ways to
>>>>>> refresh or
>>>>>> update content. I suggest there needs to be an expiration date on the
>>>>>> content just like how images and HTML pages on the web work so
>>>>>> that cached
>>>>>> content will know to pull a new copy. And if the expiration date
>>>>>> is 0, at
>>>>>> the time it was pulled, it will always get refreshed.
>>>>>>
>>>>>> This is maybe should have its own discussion thread but seems to
>>>>>> be part of
>>>>>> how this is all going to work.
>>>>>>
>>>>>> M.
>>>>>>
>>>>>> -----Original Message-----
>>>>>> From: opensim-dev-bounces at lists.berlios.de
>>>>>> [mailto:opensim-dev-bounces at lists.berlios.de] On Behalf Of Ai Austin
>>>>>> Sent: Tuesday, August 31, 2010 4:17 AM
>>>>>> To: opensim-dev at lists.berlios.de
>>>>>> Subject: Re: [Opensim-dev] Global identifiers
>>>>>>
>>>>>> myticaldemina makes a lot of good points... one thing that could
>>>>>> be problematic though relates to this comment...
>>>>>>
>>>>>>> From: <mysticaldemina at xrgrid.com>
>>>>>>> ...I would suggest any
>>>>>>> proxies would give the external system and identifier and not
>>>>>>> chain proxy
>>>>>> to
>>>>>>> proxy unless there is a reason to do it, and the assets should be
>>>>>>> copied
>>>>>> >from the original source.
>>>>>>
>>>>>>
>>>>>> I agree with the first half... no chains, just hand over the
>>>>>> external system "authority" and its given identifier pair for the
>>>>>> identity involved.
>>>>>>
>>>>>> But I don't agree at all with the idea that you then have to get
>>>>>> the asset from that original authority. The permissions could have
>>>>>> changed, corruptions could have occurred or much more likely the
>>>>>> authority simply will no longer be there. The asset "as is" (with
>>>>>> its textures, scripted content and what not) should be provided to
>>>>>> the destination location/grid if the object permissions allow it,
>>>>>> with proper transfer of the permissions to next owner exactly as
>>>>>> if an avatar to avatar transfer or rez in world took place on the
>>>>>> local grid, without trying to reload the asset from an original
>>>>>> source.
>>>>>> .
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Opensim-dev mailing list
>>>>>> Opensim-dev at lists.berlios.de
>>>>>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>>>>>
>>>>>> _______________________________________________
>>>>>> Opensim-dev mailing list
>>>>>> Opensim-dev at lists.berlios.de
>>>>>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>>>>
>>>>> _______________________________________________
>>>>> Opensim-dev mailing list
>>>>> Opensim-dev at lists.berlios.de
>>>>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>>>>
>>>>>
>>>> _______________________________________________
>>>> Opensim-dev mailing list
>>>> Opensim-dev at lists.berlios.de
>>>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>>>
>>> _______________________________________________
>>> Opensim-dev mailing list
>>> Opensim-dev at lists.berlios.de
>>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>>
>> _______________________________________________
>> Opensim-dev mailing list
>> Opensim-dev at lists.berlios.de
>> https://lists.berlios.de/mailman/listinfo/opensim-dev
>>
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-dev
>


-- 
Justin Clark-Casey (justincc)
http://justincc.org
http://twitter.com/justincc

More information about the Opensim-dev mailing list