[Opensim-dev] OGPX and IETF-ing things

[Cristina Videira Lopes]

Inifinity sent me a very nice private message, which, because it was  
private, I'm not going to forward here. But the bottom line of his  
nice message could use some public discussion. Essentially Infinity  
is suggesting that we move towards getting the Hypergrid work into  
the IETF, through this newly created OGPX working group.

Personally, I think this is all premature. IETF-ing the Hypergrid is  
premature for different reasons than IETF-ing OGP is premature. The  
Hypergrid is not really ready for prime time until we have the  
Hypergrid2 in place, with its security model to protect users from  
malicious simulators. OGP is not ready for prime time because no one  
has seen it yet, at least not any reasonably complete implementation  
of it.

The right thing to do, I think, is to first have implementations of  
both OGP and the Hypergrid2 in OpenSim. Once that is in place, we can  
all see the similarities and differences, and try to standardize the  
similar pieces. Alternatively, we can try to work together towards  
one single interop protocol, but honestly I think that's not going to  
happen, simply because there is space for many (not just 2, but 3 or 4)

What I suspect will happen is that OGP and the Hypergrid will have  
several pieces that are very much in sync, with small details that  
are different, and then they will have a few really important pieces  
that are substantially different. Things like posting/retrieving  
agents to/from regions, for example, we already converged to using  
REST; inventory access, we already converged to using capability  
URLs, etc. Small details such as the format of the data on the wire  
are different, but that doesn't really matter as long as we agree  
that the Content-type can be set to different things.

The thing that will be substantially different is the issue of  
authority: what component has authority to do what. In OGP regions  
are still the ones doing agent transfers, therefore implying a trust  
relation between interacting grids that must be established in some  
non-technical manner (i.e. the receiving region trusts that the  
sending region is not stealing the user's identity). In the  
Hypergrid, agent transfers between non-trusting regions are done on  
the client side, so that the identity of the user can always be  
verified, there is no region in the middle acting on behalf of anyone.

So, that's the main difference, as far as I understand OGP.

The Hypergrid2 is in place through a proof-of-concept prototype  
client (Grider) and a couple of small but horrible hacks in OpenSim.  
OGP implementations don't exist yet, or they are not available to us  
which is the same. I think there's a lot of work to do before we go  
and propose any of this as standards.

But I thought I'd bring this up for discussion. Maybe other people  
aren't as strict on working implementations as I am.

Crista