Some topics that did not make the previous post follow.
Interest List and…
Andrew Linden is still working on the Interest List. This is the list of things you are looking at and the viewer needs to render. New Interest List code went up on the Ahern, Balkanski ,and Beirne regions in ADITI.
One of the changes coming from the Interest List work is the removal of the server side code that runs Particle Clouds. Those are the clouds that were used in SL before Windlight came into use. So, once the Interest List code rolls out the particle clouds will be gone for good.
Removing that old cloud code will free up a few server cycles and take a small bit of data out of the bandwidth pipeline.
Also, the feed for the old V1 World Map is going to get removed at some point. Andrew expects they will shut that old API/protocol down when the new Avatar bake code starts to roll out.
Andrew is considering setting up a region in ADITI that has all the old API’s and protocols removed. That will allow testing viewers to see what breaks.
Optional Mini Map
When you play in combat games you realize you cannot hide from or sneak up on other players. The mini map gives you away.
Having the option for a region owner to turn off avatar location information could greatly improve game play. Andrew thinks it is possible. He thinks there are some mini-map specific data packets that have approximate avatar locations for avatars that are out of view. It should be possible to turn those off for a region. It would be hard to do that on a per parcel basis.
Unfortunately, there is no longer a JIRA I can send you to to vote for the feature.
The subject of large regions came up. If you play in OpenSim worlds you have probably seen them. In OSGrid, I think, we can have 2×2, 3×3, 4×4, etc. regions. May be 2×1, 3×1, and other regular patterns that make square and rectangular shapes are allowed. I’m not sure.
There are several problems with implementing mega-regions in SL. Simon Linden said, “I’ve spent some serious time looking at large regions … it’s a huge project to do it right, involving a bunch of messaging changes to the viewer (like layer data, object positions, etc), region-to-region communication (all the neighbors) our back end (the grid layout itself) … it touches almost everything in some way, which is why we’re where we are today
Open[Sim] grids emulate a bunch of regions inside one — I believe the viewer thinks it sees multiple regions, while the simulator acts like a bunch of them. It’s an interesting idea, but doesn’t scale to a large grid well.
We need to get rid of region crossings – the expectation that you can stop, serialize all your AV data, scripts, etc, send it across the country to another server farm, re-build it all, and have no lag is unrealistic.”
Andrew Linden responding to Theresa’s question about possibility of vehicle servers to eliminate the need for so much data transfer, “Yes that idea has been kicked about a lot. Theoretically possible, I still believe.”
Simon said, “We’ve thought a lot about that, but the biggest and slowest part is script data, Theresa – and that changes every frame the script runs, so it’s difficult to stay ahead of the changing state.”
So… the Lindens are thinking about ways to improve region crossings by eliminating the need for large data transfers. There are theoretical possibilities. The devil is in the details.
But, Kelly Linden points out, “Vehicles are one of the poster child cases for requiring low latency to work well, and an external ‘global resource’ running those scripts would be decidedly not low latency.”
I suspect our best bet for better region crossing is Phase II of the multi-threaded region crossing project, which we are not hearing much about these days. I think currently the building blocks it needs are being built. So, we may not hear much more until after the HTTP and Avatar Bake features are in place and working, but, that is just a guess.