SL Grid Architecture Meeting
Fleep on Sep 23rd 2007
Thanks to Tao Takashi, whom I met at SLCC07 in Chicago, I recently joined the sl-dev mailing list because I wanted to keep up with the technical changes happening in the backend, particularly since the chatter about Linden Lab open sourcing the servers has grown louder. The conversation on the list is largely over my head because I’m not a coder, but the parts that I do understand are interesting indeed.
It seems that Second Life is at something of a juncture. It’s at least on the mental map of anyone paying the slightest bit of attention to social networking, gaming, and 3d/virtual worlds, across multiple industries and sectors and countries, and there is a discernible push to open source more than just the client and to drastically improve the reliability and scalability of the grid.
These items have been hot topics of discussion on the sl-dev list, and chatting with Tao about it proved to be very helpful. He was able to explain a number of concepts from the list in a way that was accessible to me, so we talked about meeting at a set time and invite whomever was interested. The meeting was an informal rather ad hoc gathering. I mentioned it to anyone on my friend’s list who I thought might be interested in the technical details and sent out some tweets, and Tao invited some people too.
Compare the following two slides that Tao put up at the meeting(assuming it’s ok to repost since he mentioned putting the slides up on Slideshare soon):

Tao Takashi explains the current SL grid architecture model.

Proposed changes to the SL grid architecture.
In the first slide, you can see that the SL client connects to a region on LL’s grid, which then connects to the central databases. In the second slide, there are both agent (user) domains and region (sim) domains and with your client, you might connect to an agent domain hosted by someone other than LL, and then connect to a region hosted by LL or by the company or someone else. The agent domain would store one’s identity (profile, friends list, etc) and if “trusted” by another region domain, could connect the user to that region through the viewer.
This model allows for a number of interesting possibilities. Company X may host their region domain but only allow agents from within their own agent domain to connect - similar to an intranet - and may or may not allow agents to connect to other regions. Likewise, region domain hosts could set different levels of access to different agent domains depending on all sorts of criteria, but probably especially security and identity verification protocols. Trusted agent domains would have access to the most regions, and even LL could restrict certain behaviors based upon the originating agent domain (so for example preventing an agent from rezzing objects if they are not from a trusted agent domain).
It seems that now is the time to enter the conversation if you have an interest in such things, as most of the model is still very much up for discussion and debate. Soft Linden provided a number of links, as did others in the crowd, for follow-up and participation. I hope to have a chatlog soon, but in the meantime, check out these links if you’d like to participate:
- Agent Domain schema:
- https://wiki.secondlife.com/wiki/Agent_Domain
- Persistent Inventory on distributed VR systems: http://boroon.dasgupta.ch/persistent_inventory/persistent_inventory.pdf
- Information about the SL-Dev list:
- http://wiki.secondlife.com/wiki/SLDev
- Linden Office Hours and Locations:
- https://wiki.secondlife.com/wiki/Office_Hours
- Second Life Viewer Documentation:
- http://doc.daleglass.net
There was also a discussion about what this would mean to the end user, how the architecture will handle DRM, IP rights, licensing, permissions, how portable inventories would work, and how the architecture changes might impact the overall SL economy. This is certainly an important conversation and one that needs to inform whatever decision-making goes on at the code level that will enable the most flexibility, but it seems to me that those issues are incredibly complex, vary by region, custom, and RL geographic law, and happen “on top” of the underlying system architecture. My primary concern for the moment is momentum for an open sourced, scalable, reliable system, because in my mind, we’ll be arguing about the IP and economic stuff for a very long time, and I don’t want to wait that long for an open source self-hosted sim that works with the existing SL world (I am shying away from the term “main grid” purposely).
Still, those issues are important points to consider and I look forward to the discussions about how to implement a permissions system that might make it easier to license content, cite works in-world (very important to the higher ed community), and retain my inventory across regions without compromising the income of do-it-yourself content creators.
It was a great meeting and when I hear about the next one, I’ll post it here. This is my own interpretation of what was said, feel free to correct any errors or misconceptions, and many thanks to Tao for hosting and to everyone who attended and participated.
Update: Tao posted the full transcript and slides on his blog. Thanks Tao!
Filed in Second Life, Technology | 8 responses so far

