SL Project updates week 37/3: TPV Developer meeting

The following notes are drawn from the TPV Developer meeting held on Friday September 12th, and shown in the video above. Time stamps, where relevant, have been included for ease of reference to the video. Note that items are listed according to subject matter, rather than chronologically, so time stamps may appear out-of-sequence in places. My thanks as always to North for the recording.

Viewer Updates

[00:00] The Maintenance RC was updated on Friday September 12th with the release of version  This viewer includes a range of updates and fixes as specified in the release notes.

The Oculus Rift project viewer, with the first round of updates for the Oculus DK2 has been through LL’s QA process, and is currently undergoing further refinement as a result. There is no ETA on when it might appear other than “soon”; however, it is anticipated it will continue to support DK1 as far as possible, as well as supporting DK2.

As noted in part 2 of this report the  Experience Keys project viewer was updated to version and it appears further updates may be on the way.

Experimental Log-in Viewer

[21:20] For the last several weeks, the Lab has been running some A/B tests with new users using a viewer with a revised log-in screen alongside of the familiar log-in screen (see version

The log-in screen displayed by the experimental viewer the *first time* a new user runs the viewer ready to log-in to SL
The log-in screen displayed by the experimental viewer the *first time* a new user runs the viewer

Commenting on this testing, Grumpity ProductEngine Linden said:

We’ve found that there are some changes we can make that increase retention by a fairly sizeable number of thee to five percent. So we’re going to be putting forward some changes to the newbie log-in screen and then the subsequent returning user log-in screen. They’re going to be gradual, and sometimes not as beautiful as we might like all at once; however, you can be assured that they’ve all been tested and that they perform better.

The log-in / splash screen in the experimental viewer that is seen by user who have previously logged-in to SL
The log-in / splash screen in the experimental viewer that is seen by user who have previously logged-in to SL

This means that there will be an RC viewer appearing in week 38 which will have updates based upon the results from this testing incorporated into it. These changes should not affect any of the widgets  / embedded features within the log-in screen used by TPVs.

Group Chat

[03:10] The group chat work is continuing. As noted in part 2 of this week’s report, it is anticipated that there will be further tests on Aditi in the near future. However, the intimation from Oz’s comments is that something may have been deployed to the chat servers, as he indicated that the Lab were getting “good information” on recent updates.

[18:18] In reference to issues of chat servers apparently dying and requiring a restart, it appears the problem is related to a network issue the Lab has identified. A set of changes are in the process of being made to address the networks issue, and Izzy Linden from the Concierge Team indicated that he’d seen a lot fewer reports on the issue in the dy and a half leading up to the TPV meeting compared to the 7 or so days prior to that.

Z-offset Height Adjustment

[03:20] Work on the z-offset height project has commenced, although Oz underlines it is still “early stages” for the work. This is based on the proposal put to LL explaining the issues with the current hover solution implemented as a part of Project Sunshine, and describing actual usage of the “old” z-offset adjustment. It is hoped that sometime may be available “soon” with this work.

CDN Project

[03:43] The Lab has decided to change its strategy for texture and mesh fetching via a Content Delivery Network (CDN), in as much as they believe the current experimental configuration which has been under test on Aditi, and which doesn’t require any viewer updates in order to be effective, has proven successful enough that the Lab is now planning on deploying it more widely in order to see how it works under a larger load.

This means that there will be an additional deployment to a “micro channel” in week 38 (week commencing Monday September 15th). If this initial test goes well, then the Lab will likely deploy the configuration even more widely, and possibly across the grid as a whole over time, depending upon how further testing under larger loads unfolds.

If you’d like your region to be included in the micro channel testing, you should request inclusion directly from Oz. Be sure to understand and state why you’d like the region to be included and what might be tested through its inclusion.

Longer-term, the aim remains the same: the Lab will release a set of viewer updates which all TPVs should adopt reasonably quickly. This will then enable the Lab to further streamline the process of URL handling between the viewer, the CDN and their servers.

Findings from the testing to date show that in situations where you enter a region which has not been cached in your local point of presence for the CDN, then your texture and mesh download times are slightly slower than downloading directly from the Lab’s asset servers via the region’s server. This is because the CDN has to obtain the textures & mesh from LL’s asset servers and cache them locally. However, once this has been done, then texture and mesh downloads should be significantly quicker for everyone entering the region (including you, on a return visit), as the data is all cached on the CDN’s point of presence local to you.

Autobuild and Tool Chain

[25:30] The work in improving the viewer build process for Windows and Mac using MS Visual Studio 2013 and Xcode 5 respectively, together with library updates and the new viewer autobuild process (the changes for which are documented on the wiki) is close to a point where it is ready to go.  Those who like to compile their own viewer should listen to this section of the meeting for further information on this work.

Chrome Embedded Framework

[37:37] As noted in various SL project updates in this blog, and via a July blog post from LL, the Lab has been looking to upgrade the component of the Viewer that’s responsible for rendering web content, which involves implementing the Chromium Embedded Framework.  During the TPV meeting, Oz indicated this work has been temporarily stalled as those working on it have been working on the tools update project, and it is unlikely that progress is liable to be made for another couple of weeks as the tools update work is completed.

Cocoa Bugs

[38:40] The view at the Lab is that are no significant Cocoa bugs to be dealt with in the viewer. As noted in my week 31 TPV meeting report, the Lab feels fixes have been issued for camming issues, and the Keystroke entry lag problem is seen as a system issue, rather than a viewer-specific issue, and which the Lab believes can be fixed by people updating to the latest version of OS X 10.9.

For those will experiencing camming issues (such as the ALT-zoom problem), the advice is to file a new bug report on the problem and to be “agonizingly thorough” it describing the problem and steps to reproduce it and other pertinent information, as the Lab has so far been unable to reproduce issues since implementing their recent fixes.

Pipelining Viewer

[43:23] The HTTP pipelining viewer, related to Monty Linden’s HTTP pipelining work, is progressing through LL’s QA, and has been referred to as the “weaponized viewer” it is so fast as a result of leveraging the HTTP streaming.

Vivox Updates

[45:09] There is still no indication from Vivox on whether they will be supplying the Lab with an updated package for the Linux version of the viewer. However, there will be further voice updates  and fixes coming from the Lab in a viewer some time in the near future. These are current with QA, and relate more to how voice sessions are handled, rather than voice improvements, etc. For example, one of the updates should fix the issue of stalkers being able to figure-out where someone using voice teleports to in SL just by watching their voice channels.

[48:40]  This work may be combined with the Viewer Stare project, which is aimed at improving how avatars respond to voice speakers, by turning the avatar’s head to look at the last speaker. This build on the current capability for avatars to look at other typists, but places voice at a higher priority.

Avatar Awareness (aka Radar Functions)

No work has been undertaken as yet to add radar-like features to the official viewer, although it is on the roadmap for future viewer development.