The main channel did not change this week. We will likely see it get an update next week (35).
The three RC channels did get an update. This is a package that has a crash mode fix in it. Maestro thinks he can discuss this fix AFTER it rolls to the main grid.
Next week the RC channels, or some of them, will get a new package that has some visible changes.
SVC-2262 – Incorrect height value in postcard which sent from above 256m. A fix is in the coming package. Maestro describes the bug, “The bug was that the postcard-sending script did a modulo 256 on all three region coordinate components to clamp them. But, while that’s fine for X and Y coordinates, you can be >256m elevation :)”
BUG-6466 – Numbers expressed in scientific notation and include a plus sign in the exponent are not parsed as json numbers by LSL. The current package has a roll back of a previous fix to a JSON problem. So, I take it this is an improved version that will be tested next week.
STORM-2040 – Unified Snapshot floater. This got a UI update to handle this request.
Maestro tells us, “A new texture/mesh asset fetching service is up on Aditi for testing. It’s a pretty big performance improvement. April Linden has been working on it. Basically, it changes out the backend for mesh and texture asset fetches.
The viewer currently fetches all meshes and textures through the sim, which gets them from the asset server. The sim gives the viewer a ‘capability’ URL, which the viewer uses for fetching. However, with this change to use the CDN [Content Delivery Network], the sim instead gives the viewer the URL of the CDN and the CDN has hosts all over the world.
It’s good for two reasons:
- The sim isn’t burdened with texture/mesh transfers.
- You’ll often have lower latency to the CDN than the sim, which means more speed.”
April Linden said, “Because your client is establishing a connection to a server that’s (odds are) is closer to you than our data centers, you’ll get a faster response.”
Maestro, “We got pretty good results in the office…”
April, “I was pretty impressed… and we have a decent Internet connection here. :)” I assume she means at the Lab’s offices not just the USA.
Maestro, “From our office, for texture fetching, using the CDN has about double performance compared to the classic way. I measured 43 seconds with the CDN and 105 seconds with the usual way to fully load a region with many many textures after login.”
This service is live on Aditi now, in the “DRTSIM-258” channel. These regions are setup for testing:
- Region “TextureTest CDN” on DRTSIM-258 contains about 5650 textures to download, when viewed from the parcel TP location with 512m draw distance
- Region “MeshTest2H” on DRTSIM-258 contains about 5170 meshes to download, when viewed from the parcel TP location with 512m draw distance
- For comparison, regions “MeshTest2” and “TextureTest SLS” on the “Second Life Server” channel have the same content but use the ‘classic’ asset fetching method
Test comparisons show the old style mesh downloads against the DRTSIM-258 loading in 78 seconds verses 51 seconds. That is 35% faster.
Te test comparisons for texture downloads against the DRTSIM-258 loading in 105 seconds verses 43 seconds. That is 59% faster.
Europe should see a BIG improvement. So, far the Lab is measuring 2x improvement in texture download.
- Recommended test setup
- Clear cache prior to test (deleting the viewer’s cache folder is a pretty convenient method)
- Do a preliminary run before benchmarking to ‘warm’ the cache (applies to both CDN and simhost texture fetching)
- Login directly to the test region
- No other avatars around
- Draw distance set to 512m
- Maximum bandwidth set to 10,000kbps
- Viewer always kept in focus (otherwise FPS is capped at ~20fps, possibly affecting fetch performance)
- Don’t touch the camera
- For mesh benchmarking, we used the ‘Mesh Download Quiescent’ message in the viewer log. This message triggers when the viewer has fetched all the meshes it wants, and prints the fetch start and stop times (used to compute duration)
- For texture benchmarking, we used time from hitting the ‘login’ button until the time until the texture console (Develop -> Consoles -> Texture Console) indicated that 0 textures were remaining to be fetched, with “Fetch 0(0)”
- Because of the delay involved when logging in, this texture fetch benchmark is much less precise than what we have for meshes
Meastro invites you, “Feel free to try it for yourselves! (please be gentle…)”
He also explains how to tell if you are connecting to the CDN, “The easiest way to tell if you’re using the CDN is to have the viewer log show the capability URLs: Develop -> Consoles -> Capabilities Info To Debug Console. If you’re using the CDN, the GetMesh, GetMesh2, and GetTexture URLs will all be http://asset-cdn.aditi.lindenlab.com/ , with the current setup.”
The new gaming regions code is being testing in ADITI, the preview grid. The regions will have limited access due to gaming regulations. Controlling access is the point of these changes. You have to be ‘legal’ to get into gaming regions. Apparently that is a confusing issue for users right now. For what information there is see: Authorized Second Life Gaming Participants, gamblers not casino owners.
Basically you must be 19+ to get in and have payment info on file. Also, ‘participants’ must have a RL address on file.
There is a gaming region setup now on the main grid named “Crunchy”. There will be no games there. It is a test of the access control.
There was more pile-on testing of changes being made to the group chat system. The current change delays the member list download. The list download will be cued. These changes only kick in for larger groups. We may see these changes move to the main grid if they prove themselves on the ADITI grid.
The main viewer has updated to the Library update, 3.7.14-292638.
As of today, Friday, there are no RC viewers listed.
Experience Tools Viewer version 188.8.131.521846 – we know what this one is.
OculusRift Viewer version 184.108.40.2062141 – This is the DK1 test version. CtrlAltStudio has a preliminary viewer version out for the DK2. The Lab is working on a revision to adapt to DK2.