The following notes are drawn from the TPV Developer meeting held on Friday, October 24th, 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.
[00:10] There have been a couple of end-of-week viewer updates:
- The HTTP Pipelining viewer was updated on October 24th to version 220.127.116.115700, incorporating the most recent viewer releases
- The Benchmarking viewer (which removes dependencies on the GPU table) was also updated on October 24th, to version 18.104.22.1685759, incorporating the most recent viewer updates.
Both of these viewers have performed well both in the RC role (HTTP Pipelining) and project viewer (in the case of the Benchmark viewer), and it is likely one or the other will be promoted in week 44 (week commencing Monday, October 27th), although the promotion may not occur until later in the week, depending on how these new releases perform.
[01:58] The Experience Keys project viewer is in the “final throes” of bug fixing. However, it is unlikely to appear as an RC viewer until after the next set of server-side Experience Keys have been deployed. When this will be is unclear. It is not “imminent”, but by the same measure, it is not “distant”.
[02:33] The Oculus Rift work continues, but again, this is unlikely to move beyond being a project viewer until such time as the Oculus hardware is more generally available. The Lab will continue to keep the viewer code base current with releases, and will doubtless continue to test and refine as further SDK updates appear (such as the just-released 0.4.3).
[03:16] Progress is continuing to be made with the viewer-managed Marketplace updates (VMM). There is now a technical wiki page for the web API available, and Skylar Linden was on-hand to provide and overview of the page, which by his own admission, is “pretty dry stuff”, but includes things like:
- Figuring out if a user is a merchant or not, and how the system will respond (loading the correct page or not)
- A data dictionary that gets transmitted back and forth between the viewer and the Marketplace
- Links to obtain listings, create listings, modify listings, associate inventory with listings, and delete listings (which are the Marketplace functions specific to the new capability).
[08:40] The ability to delete listings is a feature that is new to the Marketplace as a result of VMM, although it is somewhat confusing in that – as Brooke Linden explained in the meeting – it doesn’t actually delete anything, but gives a means for merchants to remove listing information by making it unavailable. As such, listings deleted in this way will be non-recoverable.
[10:03] The Lab is now “very close” to getting the VMM updates available on Aditi’s Marketplace. When this happens, testing will initially be with a “small group” of merchants and TPV devs using a project viewer, the code for which will be made available as the testing starts in order to allow TPVs to integrate it into test versions of their viewers. If all goes according to plan, this initial test could start in the next couple of weeks, with the aim of getting initial feedback.
Once this has been done, testing will by opened-up to allow wider involvement. This is expected to happen “within the next couple of months”, and will be aimed at allowing anyone who would like to try VMM. There will be a beta testing application form made available for those interested.
[13:45] The Lab have been continuing to roll changes to group chat, and the thinking is that the results of the changes are “pretty good”, and the Firestorm support team are reporting they’ve seen an improvement as well. As noted in part 2 of this week’s update, the work isn’t finished – Simon is looking at the possibility to improve code elsewhere in the system; there is also liable to be a round of code clean-up as well. However, Oz cautioned that the Lab may have implemented the changes which may have the biggest noticeable impact in terms of improving the group chat experience.
[14:49] Again, as noted in part 2 of this week’s report, the CDN is now supporting all the primary Server RC channels. so far, the results have been “excellent”, but the Lab is waiting to see how things fair over the weekend, when the grid is liable to see its heaviest use in terms of numbers of users on-line since the RC deployment was completed this week.
The load on the sim host Apache services is described as “way, way, way, down”, which is a good thing, as it means that all the other things the Apache services have to do (such as participating in region crossings) are no longer being impacted by the simulators handing the texture and mesh fetching loads.
One offshoot of the CDN work is that the Lab is likely to spend time validating whether the viewer cache is working as well as it could be, even allowing for the interest list updates. There are apparently differences in how well the cache works across different code-bases, so this is something that TPVs may well be involved in as well.
Z-offset height Adjustment
There has been no further progress on this since the last TPV meeting.
AIS v3 issues
[19:16] As noted in the last TPV Developer meeting, the Firestorm team pointed to a series of issues potentially related to the deployment of AIS v3, including the fact that a fix for one of the issues, as included in the (then) current Maintenance RC had spawned two further issues (the Maintenance RC was subsequently withdrawn by the Lab). Six issues in particular were highlighted by the Firestorm tam, and the Lab has been attempting to investigate / address them:
- BUG-6197: this is now believed to have been fixed, and is in a new Maintenance viewer which is currently held-up due to the discovery of an unrelated showstopper bug
- BUG-6925 and : these last two issues may be the result of race conditions arising from the timing of certain events handled by the back-end services being changed at the time AIS v3 was being deployed. The Lab is continuing to investigate both, but in the case of BUG-6925, they have been hampered by an inability to successfully reproduce the issue, which they are still trying to address.
[27:07] Users have been experiencing problems with voice breaking-up when using viewers running Vivox 4.6. A series of user-led tests were carried out, which resulted in a comprehensive bug report on the matter being filed with the Lab (BUG-7556), together with a video demonstrating the issue, and comparing the use of Vivox 4.6 with the earlier 4.5.
As voice is not something the Lab can address, the bug report was passed to Vivox, who were (as with the Lab) impressed with the thoroughness of the report, and are now investigating the problem.
As an aside to this, Oz Linden indicated that the Lab will be working with Vivox “over the next couple of quarters” looking at issues of voice quality and reliability.
In addition, and while it has yet to be confirmed as something which will be done, the Lab is considering altering how voice is packaged with the viewer – making it a .DLL that runs within the viewer, rather than a separate .EXE, as is the case now. However, if the Lab opt to do this in the future, and assuming they are not by then building their own 64-bit viewers, TPVs producing 64-bit versions of their viewers will need to continue to use the .EXE voice plug-in (which will remain available).
Tool Chain Xcode 6 Frame Rate Issue
Those building the Mac version of the viewer using the new tool chain set implemented by the Lab are reporting high frame rate drops on versions of the viewer built using Xcode 6. These issues have been noted by the Lab, but they are not yet in a position to start addressing performance issues within the new builds, as they are still attempting to get the build process to work uniformly. However, there are no plans to start releasing viewers built using the new tool chain until such time as this issue can be investigated and hopefully addressed.