The following notes are primarily taken from the TPV Developer meeting held on Friday, April 24th, a video of which is included below (my thanks as always to North for recording it and providing it for embedding), and from the Server Beta meeting held on Thursday, March 26th. Any time stamps contained within the following text refer to the TPV developer meeting video.
Server Deployments, Week 17 – Recap
As always, please refer to the server deployment thread for the latest updates.
- On Tuesday, April 20th, the Main (SLS) channel received the server maintenance package deployed to all three RC channel in week #16, which comprises internal server logging changes and new flags for llGetObjectDetails()
BJECT_BODY_SHAPE_TYPE– returned list entry is a float between 0.0 and 1.0. Anything > 0.5 is male, otherwise female; -1.0 if the avatar is not found
OBJECT_HOVER_HEIGHT– returned list entry is a float, -1.0 if the avatar is not found.
- There were no deployment or restart on the three RC channels on Wednesday, April 22nd.
SL Viewer Updates
[05:50] The Tools Update viewer, version 188.8.131.520918, was promoted to the de facto release viewer on April 23rd – see my article here for details. During its run as an RC viewer, this release had around a 2% lower crash rate than the release viewer built using the “old” tool set and processes.
As a result of this, all the remaining RC and project viewers are being updated to match the release viewer code base, and updated versions should be appearing soon.
Attachment Fixes Viewer (Project BigBird)
[07:42] This viewer, current available as a project viewer – version 184.108.40.2060856 – and which fixes a range of issues related to avatar attachment failures, is in the process of being updated to a Release Candidate status, and should be appearing in the release pipeline as such in week #18.
[29:28] The Lab believes that these fixes resolve all of the viewer-side issues related to attachment problems which are related to AIS v3. However, a number of the more noticeable issues – such as problems with attachments being detached on teleporting – are server-side, and require further investigation / fixing. Similarly, failures with requests to attach multiple items (such as during an outfit change) also appear to be simulator-related, rather than anything within the viewer or linked to AIS v3.
Oculus Rift Viewer
[07:55] The Oculus Rift viewer is now on the schedule to be updated and brought into line with more recent viewer code releases. There is no set time scale for this project (and the Oculus Rift itself, according to Oculus VR, is unlike to reach a consumer release in 2015), but the aim is to bring it back to a “more active” state.
[00:08] Brooke Linden gave an update on VVM – as this is of interest to a potentially wider audience than those interested in viewer development, I’ve provided a separate article on it.
Web Media (Webkit and CEF)
[08:41] The Lab is making “pretty good” progress on replacing webkit, an increasingly outdated third-party library used within the viewer for powering the built-in web browser, displaying web profiles and powering in-world media (TVs, MOAP, etc.), with the Chromium Embedded Framework. The Mac work is lagging a little behind this, but the Lab has now called-in external expertise to help move the project forward as a whole.
Request for TPV / Open-source Support for Linux
[09:17] The Lab is seeking support from TPV developers and the open-source community to help maintain and move the Linux flavour of the viewer forward. For details, please see my separate article in this blog, complete with an audio extract from the meeting.
Snapshots to E-mail
[12:27] The send snapshot as e-mail capability is in the process of being removed from the viewer.
The main reason for this is that wherever possible, snapshots are sent via the “secondlife.com” domain, but use the sender’s own e-mail address as the originating address in the “from” field of the sent e-mail which appears as if the “from” address is being forged. This, and other ways in which e-mails flowing out from “secondlife.com” are handled, has resulted in some ISPs regarding the domain as a spam domain, and have been pro-actively blocking it (Germany-based GMX is one such example).
To rectify these problems, the Lab is reviewing how e-mails from “secondlife.com” are being managed as a whole, and eliminating those uses which may conceivably lead to the domain In the case of the snapshot floater, the Lab’s perspective is that the easiest way to fix the problem is to remove the option from the snapshot floater; however as was pointed out to them in the meeting, this will break content such as wardrobe HUD systems which utilise the snapshots to e-mail functionality.
HTTP and CDN Use Expansion
[20:35] The Lab is working on increasing the number of assets such as animations, sounds, and gestures, consumed by the viewer to being delivered via HTTP the CDN, and removing the reliance on UDP. This is for a number of reasons:
- It further fees-up resources on the simulator to do what they do best – simulate the world around us, rather than using them for managing UDP file transfers
- The use of UDP is not the most efficient or robust means of carrying out these transfers
- UDP is bad for the network; there’s no flow control packet or congestion control behaviour, it can result in high packet losses which may occur anywhere between the the server and the viewer, and thus be hard to identify and prevent in future, etc.
As this work progresses, the Lab will be removing the server-side support for the UDP messaging currently used by such transfers. This has already happened with inventory fetching (and the option to disable HTTP Inventory is due to be removed from the viewer), and will be happening soon with texture fetching (which will also see the removal of the option to disable HTTP Textures in the viewer).
To help with this, TPVs are being encouraged to work with the Lab to identify specific / reproducible issues users are encountering vis HTTP, etc., so that more work can be put into fixing them, and the Lab is asking TPVs not to recommend to users to switch back to the “old ways” of doing things when potential HTTP problems are encountered, as the 2old way may not be around for much longer.
[27:50] There will be a “bunch” of server-side inventory fixes being deployed soon which are specifically aimed at eliminating a range of inventory loss issues, and which have been implemented directly as a result of the recent inventory loss survey people were asked to complete. Some of the server-side issues may well be related to CoF issues which can occur, such as BUG-8944.
Inventory Cache Study
[35:24] The Lab would be interested in supporting any TPV developer who would like to undertake a deep study of viewer caching with a view to identifying issues and problems which may occur with it.
[38:04] As noted in my last TPVD update, having large “flat” inventories (e.g. tens of thousands of items within a single folder, rather than being split between multiple sub-folders) can cause the log-in process to slow significantly for it it to time-out and the simulator to disconnect the viewer.
To help alleviate this, the Lab will be deploying an inventory transform which can be run by support against accounts experiencing these problems, and which splits-up the items within the “flat” folder(s) by date, and places them into sub-folders.
In addition, the Lab is considering updating the viewer so that if the contents of a folder exceed a certain threshold (possibly 5-10K), it will display a warning suggesting it might be better to split the items into sub-folders.
Teleport Failures and Ban Lists
[43:33] The Lab’s support staff have seen an increase in teleport failures which are the result of regions having large ban lists at both the region / estate level and the parcel level. This causes extra delays when attempting to teleport into a parcel within the region, and can result in time-out. Wherever possible – and the Lab acknowledges this cannot be done in all causes – the preferred route to to keep the ban control at either the estate / region level or the parcel level.
While not necessarily directly related to this, the Lab plan to have a further look at the back-end of ban lists in the future.