[Opensim-dev] Global identifiers

diva at metaverseink.com diva at metaverseink.com
Sat Aug 28 23:36:14 UTC 2010


Dahlia Trimble wrote:
> Would it be possible to use just the UUID part and not tie it to a 
> specific domain? I can foresee events where a user and a 
> business/grid/domain owner/whatever make a decision to go separate ways 
> and then any other systems out there that have the global identifier 
> referenced may not be able to use it to contact that user. If only the 
> UUID were used (we already use UUIDs for many diverse applications and 
> the context implies the specific usage) it could be similar to a cell 
> phone number where a user can take it from carrier to carrier. Resolving 
> the current user domain would need to be worked out but several methods 
> might be considered from a P2P network of resolvers to one or more 
> central authorities, and/or simply storing it in a cache in local 
> installations when the UUID is first used.

This is an interesting idea. Two points:

1) In the unregulated S2S system we have at hand, the uniqueness of 
UUIDs cannot be guaranteed. Just like on the rest of the internet: 
anything "inner" than domain name is not guaranteed to be unique. The 
Hypergrid is designed along the design lines of the Internet, and it 
doesn't try to impose any more constraints. Same thing with archives: 
Nebadon posts his OARs on the web, and there's no saying to where they 
will end up.

2) Because the way the Internet works is so general, there's nothing 
preventing groups of grid operators from coming together and 
establishing protocols that enable new value-added services, either just 
within their group or between their group and the rest of the world. 
Like, for example, ensuring that users' local identifiers (UUIDs) never 
go stale in between them. As long as the user switches from one member 
of the same group to another, the local UUID will be resolved within 
that group. This can be done in many ways. I'm personally not interested 
about the details of those smaller arrangements (they seem technically 
trivial, when there is law-enforced trust), but it's important that they 
can exist over the more general protocols that we are designing here. 
They can.




More information about the Opensim-dev mailing list