[Opensim-dev] Revising the grid standards (Part II -Implementation details)

Bowman, Mic mic.bowman at intel.com
Fri Nov 2 19:55:17 UTC 2007


i agree with sean re: xmpp (i have done quite a bit of programming with
large scale eventing systems based on xmpp)... for some things the
current set of implementations are really good (eg *typed* IM's) but it
is very hard to get the performance you want out of most of the
implementations. the handshake protocols required to set up message
dialogs can be pretty heavy. (but i'm VERY happy that we would consider
XMPP for instant messaging)

having said that... could we also open the discussion about what set of
functions is inherently part of the architecture & what are peripheral?
For example, IM/chat could be separated from the VW functionality
assuming the VW architecture exported some reasonable set of event apis
like "enter local region", "changed region" and so on. Are there other
pieces where the focus is integration of existing systems?

--mic

 

-----Original Message-----
From: opensim-dev-bounces at lists.berlios.de
[mailto:opensim-dev-bounces at lists.berlios.de] On Behalf Of Sean Dague
Sent: Friday, November 02, 2007 12:01 PM
To: opensim-dev at lists.berlios.de
Subject: Re: [Opensim-dev] Revising the grid standards (Part II
-Implementation details)

On Sat, Nov 03, 2007 at 12:26:30AM +0800, Adam Frisby wrote:
> Following on from the previous grid standards email, we should now
look 
> at what we need to revise and propose some actual hard specifications 
> that we can use.
> 
> Some base guidelines:
> * No EventQueueGet or anything similar. If it needs speed, make sure
we 
> can specify something like a remoting serialiser and use TCP sockets
for 
> this.
> * Re-use existing standards and libraries where possible. No creation
of 
> unnessecary standards without good justification (ie LLSD vs JSON)

++++++++ (I think I've run out of pluses.)  Can't agree enough on this
point.

> MW has suggested we look at using XMPP for this - from my cursory 
> examination this looks like a reasonable protocol that we could
actually 
> use for at least things like handshaking and core [inter]grid
> protocols.

After spending some time tracing on Tleiades AssetServer, based on REST,
REST has my vote.  The flexibility to get data in and our of the
environment is really nice.

The other nice things with REST is that HTTP load balancing techniques
are fairly well understood in the industry.  For instance, there are
even techniques and products for session aware HTTP spreaders.  It also
has the advantage of being similar to the aproach Zero has stated the
Linden grid is moving in, so future interop with the Linden grid will be
easier if our base services are REST.

That being said, I'd really like to see XMPP integration for IM if
possible, as it will make for easy bridging of in world presence to out
of world IM systems.

> The core grid functionality (what needs defining - this list will need

> to grow, currently just including things we use)
> 
> -- What needs defining --

I assume all the User & Agent services are Region <-> Grid Server(s)?
Please correct me if I'm being dumb here.

> * User Services [grouped together]
> ** Profile Retrieval
> ** Instant Messaging			--> Move to client?
> ** Inventory Handling			--> Move to client?
> 	* List Inventory		--> Move to client?
> 	* Rez Inventory Item		--> Move to client?

Can you define move to client? in this context?

> ** User Login to Grid
> 
> * Agent Services [grouped together]
> ** Session Retrieval
> ** Session Authentication

Both of those are Region <-> Agent Server for user info?

> ** Region Login from Grid

This is the circuit setup?

> * Region Services [grouped together]
> ** Region Information Retrieval
> ** World Map Services
> ** ID/Position to URI translation

I'm not sure I fully understand what all the functions in each of these
are, can you explain a bit more?

This is a really great post, btw, in clarifying what it is that we need
to start defining.  Thanks for getting this out here Adam.

     -Sean

-- 
__________________________________________________________________

Sean Dague                                       Mid-Hudson Valley
sean at dague dot net                            Linux Users Group
http://dague.net                                 http://mhvlug.org

There is no silver bullet.  Plus, werewolves make better neighbors
than zombies, and they tend to keep the vampire population down.
__________________________________________________________________



More information about the Opensim-dev mailing list