Lab announce Project Shining complete

On Monday June 16th, Linden Lab issued a blog post stating Project Shining is now complete. This is a major milestone for the Lab, with Shining representing some 2 years of effort and has involved significant work in three key areas:

  • Avatar baking (Nyx Linden) – also referred to as server-side baking or server-side appearance (SSA), which comprised a major project to shift the heavy lifting of avatar baking / appearance from the viewer to dedicated baking servers. for those need to know more, I offer the post I wrote back at the end of December 2012 as the project saw the release of its first viewer. Further information can be found under the Server-side Appearance tag
  • Interest List and Object Caching (Andrew Linden) – a project to improve how scene rendering is handled by both the server and the viewer, and improving the viewer to retain more information on regions in its cache and re-use that information more intelligently by the viewer
  • HTTP updates (Monty Linden) – a major update to SL’s communications infrastructure to make more and better use of HTTP in order to improve viewer / server communications and increase their reliability and their efficiency.
The state of play with some of the HTTP work in May 2013
The state of play with some of the HTTP work in May 2013

Each of these sub-projects have comprised various stages and releases. Interest List and Object Caching, for example went through several rounds of updates alternating between server-side work and viewer-side work, with the final round of work, focused on improving how the viewer caches and reuses information, reaching a release status in May 2014.

The Project Interesting video released by the Lab in May 2014

The reason for the Lab’s announcement about project Shining is that Monday June 16th saw the release of the final set of Project Sunshine viewer-side updates in the form of the new de facto release viewer, version:, which also includes significant inventory updates and improvements which have been referred to as AIS v3 (Advanced Inventory Service version 3, although this requires additional deployment of server-side updates across the grid). This viewer also includes a number of viewer memory leak fixes as well.

While the implementation and deployment of some elements of the work did experience some hiccups, overall, each element of Project Sunshine was implemented very successfully and with little or no disruption noticeable to users. Each has already yielded significant improvements to both the overall service and in the user experience, and the Lab are to be congratulated in bringing all three of the Project Sunshine activities to a successful conclusion. Here’s to the next round of projects – such a HTTP pipelining!

9 thoughts on “Lab announce Project Shining complete

  1. I thought I had stopped hearing adverts like that when Pearl & Dean retired its restaurant advert. You know, the place so good that even the chef eats there.


  2. Mmh. Just tested the new version. It has precisely the same issue that have prevented me from using any recent version in the past two years. Since LL switched to Cocoa, text chat delays make conversations impossible — it’s a long-standing bug which affects a considerable majority of Mac users, but sadly not all, or someone would have found a fix by now. Allegedly the latest alpha version of Firestorm has fixed the major bugs by now, but I’m still waiting until it gets ‘officially’ released. Until then, I’m back to an early 2013 version of Firestorm (pre-Cocoa), which is the only way I can still enjoy SL.

    Granted, the tests I just made showed that LL has done a slight improvement towards fixing this issue. Typing in real time seems to have much shorter delays than before. Later on today, I should be on a very chat-intensive meeting, so I’ll see how bad it gets after a few minutes 🙂

    But I’m not having high hopes. The JIRAs for that on both the LL JIRA and the Firestorm JIRA continue to get regular updates from Mac users, telling how nothing has improved, and communication about that particular Cocoa bug has been not very profitable. LL in particular assumes that the bug doesn’t officially exist. The Firestorm developers at least acknowledge its existence, but they shrug it off as part of the Cocoa implementation, and, as such, very hard to fix… compared to doing all the rest which needs doing!

    Oh well. So long as I can still log in with the old version of Firestorm, I’m fine. It’s a pity, really, because the more recent versions have such a boost on performance…


    1. Cocoa and other long-standing Mac issues (such as the ALT-Cam problem) weren’t a part of Project Shining, but are being slowly dealt with via other resources at the Lab.

      The problem here is two-fold: the nature of some of the issues, which the Lab themselves admit they’re having difficultly getting to grips with; and resources. A lot of people at the Lab use Mac systems, but are not themselves Mac developers, so the pool of skills they have is small (TPVs are actually faced with the same issue as well). On the sort-of upside of this, more effort is going into trying to clear those issues which do have a relatively straightforward fix and get them out: there have been a number of such fixes in viewer releases over the past several weeks.

      Sadly, the bigger issues are going to take time to tackle – I know that’s no comfort to you or anyone else suffering under them, but the problems aren’t being ignored. They are routinely discussed and thoughts / ideas offered at things like the Open-source Developer meetings and the TPV Developer meetings, TPV devs have been trying to help where they can, etc. But the unfortunate situation is, no-one seems to have a clear answer on overcoming / resolving the problems.


  3. As i don’t use linden lab official viewer and the Tpv’s i use did not update yet to the latest code, i can only tell what i experiencing via server side changes.
    Much less crashes when crossing sims, that is some i noticed already in the last months.
    But since a few weeks ago, i notice that the rezzing time is taking a lot more then usual unless…
    fix the camera on a subject, lower the draw distance to 32m, let it rezz fully, then adjust the draw distance to what you wish.
    I do believe this effect is due to the viewer side code is not yet implemented.
    And ‘m eager waiting for the time where all users will have their viewers with the viewer side code in place.
    Still can any tel me if the mouse lag issue that some are reporting after adding the latest code to they viewer builds and im not speaking about mac users but window or linux, is happening also on this latest LL release? Cause if i read correctly it makes almost impossible to use the combo crtl +alt ledt mouse to pan and zoom around, like i do.


    1. I tested the last LL viewer on Linux (I didn’t check on Windows) and there is the same super slow response when I use alt+click to move the camera. It happens the same with the global map (Ctrl+M), when I try to click and drag the map. It feels like mouse sensitivity goes wrong.


      1. Yeah I get this too, though I noticed the higher I turn up my dpi on my mouse it seems to fix it a bit, though I used to use 400dpi at 1920×1080 which some people might see as considerably slow to begin with.


  4. Ok, it is really some that is dully broken as well on windows!
    God, i do hope that LL can really fix this asap. I never used the camera movement floater and cant even wonder how i wil manage to be in world if i have to use that thing to all!


  5. Quoting from Niran’s Nv blog:
    . This update also includes the recently released “Interesting Project” from Linden Labs so expect bugs and a really laggy direct avatar camera movement (sadly :/, i hope they fix that).

    So sadly it real looks like Linden Lab screwed some that is essential, for me at least.
    I do hope that they realize how important this feature is.


Comments are closed.