This last Friday was the Third Party Developers’ meeting facilitated by Oz Linden. We have a burst of news from that meeting. Several projects were updated.
I’ve already covered the update of the SL Default/Main Viewer and the RC Viewers. There was no new news regarding these viewers.
However, some props are due NiranV. The new snapshot panel is his creation. It is well done. I personally like using it. NiranV also corrected some other problems in the snapshot process. So, it is more than just a cosmetic change.
Send NiranV a thank you.
The Oculus Rift Project viewer is in QA as a DK2 project and lots of issues are being found and worked on. Fixes are flowing into that project viewer. So, far the viewer will support DK1 as best it can.
My understanding is the DK2 version of the project is NOT yet published on the RC Viewer page, but soon.
The Experience Tools Viewer has been updated with a new round of fixes and is getting more. That Project Viewer will soon get another revision.
A project probably to be known as the Pipelining Viewer is coming along. It is reported as being ‘screamingly’ fast. QA people have referred to it as the weaponized viewer. This is the viewer that uses HTTP streaming, something Monty Linden has been building toward for months with his HTTP work.
A set of voice software changes are coming, which may make it into a coming RC Viewer named Viewer-Stare, real soon. They include an anti-stalking measure related to voice. The real point of Viewer-Stare is having the avatar look at the current speaker using voice. This is an expansion of the feature that causes the avatar to look at who ever is posting in chat.
There is an Awareness Project… several developers were unaware of it. So, was I. It has to do with our mini-map or RADAR. See: Useful Radar Options. It appears this has made the list of things to implement.
The Awareness Project is a plan to create a better RADAR with more features. I’m not sure how well that may go over with the combat game managers, I suspect players will love it. But, that will pretty much annihilate any possibility of a sneak attack.
Work on Group Chat is proceeding. Oz says lots of good data is being collected, which will guide changes they are making.
You have probably noticed group chat servers apparently dying. The servers are not dying, they are running out of ports. Recent changes have been made and will soon or have already been deployed. So, we should see some relief. But, the Lindens are not certain.
Other possible changes to chat are being discussed within the Lab.
Support is getting better at restarting chat servers. So, time from request to restart is shrinking.
You know the Hover setting in Appearance is not all we want it to be (search here on HOVER if you don’t know about the problems). This project is an effort to correct those problems and provide a better solution. My klutzy solution is here: Second Life Hover Solution.
The Hover Project is looking at an elegant way to correct the problems associated with avatar positioning, both standing AND SITTING.
Oz tells work is proceeding on the project.
This projects moves texture and mesh downloading off the region servers and onto commercial Content Delivery Networks. Testing has shown the system to provide significant performance improvements.
Oz says they have changed their thinking on how to implement the newer system. For now they are going to roll out the CDN system that requires no viewer changes. They will watch how it performs and make adjustments based on results.
The change will roll out on a micro-channel made up of AGNI servers/regions during week 38. Oz says they will be creating a NEW micro-channel or 16 or 32 regions as current channels are larger than they want to work with.
If you want your region in the test channel, contact Oz Linden. Include a reason for why you think CDN will help you or what it is you plan to test.
At this time there is no information on which regions will be selected or how they will be chosen. I would expect a collection of regions ranging from empty and deserted to full and crowded. I think that would give the largest range of interesting comparisons.
The change is already running on some regions in ADITI.
There are no known viewer compatibility issues. So, it should just work. Oz, acknowledges those are often spoken last words.
If all goes well, this change will go grid wide. The Lindens plan to do a slow roll out testing as they increase the number of regions using CDN. If all goes well, the entire grid will move to CDN.
Timing? There is no ETA for this slow roll, other than the micro-channel starts next week. What is done next depends on results in that channel. I’ll speculate that if all goes well, this could be grid-wide in 3 to 6 weeks. Of course if things do NOT go well, it will predictably take longer.
If this slow roll makes it to the entire grid, the next step will be changing part of the viewer. As it is now the regions give the viewer a URL to use for requesting assets. The viewer modifies the URL to add the asset being requested and sends it to the region, which then modifies the URL to create the final URL. I think this is returned to the viewer which then makes the request from the SL asset system.
The plan is to remove the region URL-modification process. The viewer will be given a CDN friendly URL by the region or possibly the login servers (?) and the viewer build a CDN friendly URL and asset ID. The asset request will go to the CDN servers closest to you and CDN will deliver the asset. Changing to this process will require the viewer to change.
This shortens the pipeline and will improve performance.
The textures and meshes coming from CDN will have cache headers and use port 80, just as web content does now. This will allow not only the CDN to cache the assets, but also allow ISP’s and others along the pipeline to cache also. So, you may some additional benefit from local ISP caching.
The current system uses a URL with a query string. You have seen query string in URL, but you may not recognize them for what they are. In a YouTube request below the query sting is the red part.
The problem with this type of URL is that some caching systems ASSUME such URL’s are not cacheable. If your ISP is one of those, you will lose the benefit of your local ISP cache. When the viewer is eventually changed the URL will avoid the use of a query string. Then your local ISP cache will kick in and help.
The change will significantly reduce the load on SL servers. It will also reduce the benefit of users running SQUID caches on their local machine.
There is some question wither the CDN can cache a data set as large as what Second Life uses.
Login Screen Changes
The A-B testing of the login screen changes has shown fairly large improvements in player retention for changes to the login screen that will soon start appearing.
We will see an RC Viewer with the changes come out in Week 38.
If you dig in Bitbucket as some do, watch for viewer-release-nui-login.
The Lindens are encouraging Third Party Dev’s to implement the same changes in their viewers. It seems some number of new users make their first login using a things party viewer, which I find a bit surprising. I would love to know what the numbers are for that aspect of viewer use.
This project only indirectly affects most SL users. The auto-build is about how viewer software is built. If you are not compiling a viewer, these changes have little interest for you. But, they will improve the viewer you use.
For more information on the project see Oz Linden’s User Page.
A next step is to upgrade to VS2013, a development and compiler tool that Microsoft makes.
For more information listen to the meeting recording starting at about the 28 minute mark. If you are going to compile a viewer, this is almost a must hear.
The Lab is hoping a third party developer will pick up the work to provide the new SpaceNavigator v10 support, as it will be some time before the Lab gets to it.
Techwolf Lupindo brought up the issue in the meeting.
Sine I wrote about my trying to get the SpaceNavigator v10 software working, I’ve heard more about people running into the install problem. Some just omit installing the driver and uses Window’s default driver, which is a workable solution for SL use.
The Lab is replacing the WebKit with the Chrome Embedded Framework (CEF). That work is on hold while the Lab gets the auto-build tools project completed and then re-compiles all the libraries associates with the SL system.
Once that work is complete the people on that project can return to working on implementing CEF. That is expected to happen in a couple of weeks, so maybe week 39 or 40 work will resume.
Oz believes there is nothing left for the Lab to do with these issues. I think this means that most of the remaining problems are not so much SL Viewer problems as they are Mac problems. There are other Mac changes they will eventually get to.
The Alt-Cam bug cannot be reproduced by the Lab. So, if you are experiencing the problem file a JIRA and document it in GREAT DETAIL. Otherwise, there isn’t anything for the Lab to sink its teeth into and Oz won’t assign people to it.
The slow typing bug is fixed by upgrading the Mac OS to 10.9, a free update.