SL project news: week 5 (1): servers, viewer, deformer

Server Deployments Week 5

There are a full set of deployments this week, Main channel and RCs.

On Tuesday 29th January, the Main channel received the threaded region crossing code deployed to BlueSteel and LeTigre in week 4. This project makes sim performance smoother when objects and avatars cross between regions. Please refer to the release notes for further information.

On Wednesday 29th January, the Release Candidate channels should receive the following deployments:

  • BlueSteel should receive the server-side materials processing code. This code will require a a project viewer in order to be used, As reported in week 4, there is no definitive timeframe as to when a viewer is liable to be available, but it would be reasonable to expect something in the next few weeks (possibly sooner) – release notes
  • LeTigre should receive the maint-server project originally deployed to Magnum in week 3. This deployment includes the bug fix for the places / search indexing issue which occurred following the original deployment of the code to Magnum – release notes
  • Magnum will receive a further update to the interest list code deployed in week 4. This update addresses high packet loss problems, leading to issues such as not seeing avatars or objects in-world, specifically:
    • A fix for bots which do not specify a valid draw distance in order to prevent them downloading huge amounts of unwanted data
    • A fix to keep ObjectUpdate packets within an ethernet MTU of 1500
    • Release notes.

Please check with the forum discussion thread for these deployments to keep abreast of updates / changes / issues.

Materials processing: server-side code to go to BluseSteel on Wednesday 30th January - but no project viewer yet
Materials processing: server-side code to go to BlueSteel on Wednesday 30th January – but no project viewer yet

SL Viewer News

Problems remain with the beta viewer code, but it is hoped a new beta will be appearing shortly. In te meantime, Kelly Linden has committed the viewer-side code for extending the maximum avatar animation length from 30 seconds to 60 seconds (MAINT-1492). This should be appearing in the 3.4.5 beta viewer, and should also be appearing in TPVs in the near future.

Mesh Deformer

Delays continue with the mesh deformer, which can be summarised as:

  • The Lab has limited internal resources for testing it – although attempts are being made to address this
  • There is concern that there is no clear way for a designer and a user to be sure that they are using the same avatar base model – and if they are not, then the deformer will very likely make things worse. While there are some provisions within the current deformer for the base model to be taken into consideration, there is apparently concern that it may not be sufficient
  • The project veiewer code is now significantly behind the current viewer release, although work is underway to merge it up (and I received a report that this has been done while preparing this update).

A further potential issue in progressing the deformer  – as Karl Stiefvater (Qarl Fizz), who wrote the original code points out – is there are a lot of expectations that the deformer will do a lot of things which were not in the original specification (or indeed, the code as originally written). Some of this scope creep has resulted in further updates to the code from other contributors. However, there appear to be concerns that if the Lab prematurely release if, there will be a risk that a) it breaks existing content, b) it is viewed as not addressing problems correctly.

This latter aspect was also commented upon by Oz Linden, when speaking Open-source Development meeting on Monday 28th January. He said, “We are often (and frequently with justification) accused of throwing things out that are not really finished enough… indeed, the lack of some solution for the problem the deformer is trying to solve is one of the more prominent recent examples, but I don’t want to make the situation even worse by putting out a ‘solution’ that turns out not to solve enough of the problem and/or creates bigger ones down the road.”

Avatar Baking (SSB)

A set of code updates for the viewer are in progress. However, an issue with the Mac build means they have not been pushed as an update to the project viewer as yet. It is hoped that the updates will reach all three versions of the viewer (Windows, Linux and Mac) once the Mac build problem has been resolved.

6 thoughts on “SL project news: week 5 (1): servers, viewer, deformer

  1. I cannot avoid the thought that bad communications by Linden Labs have been adding to the problems with the Mesh Deformer. We shouldn’t have to depend on people such as yourself or Nalates to draw a coherent picture of what is happening. (For various reasons, Oz Linden does not impress me.)

    The MC Rollout, going by the Status announcements, seems to have been unduly prolonged. There have been some reports of a problem occurring while the Region is caching, and I saw some strange things immediately after a region restarted which were consistent with that. I had controlled movement in AV flight with out-of-range locations in a neighbouring region, rather than stopping at the boundary or changing region.

    When I later did a test with a vehicle, sim crossing seemed perilous, but the problems were limited to rubber-banding. There were no crashes, no sudden heading changes, a generally more reliable, if flawed, experience.

    Once the RC rollouts are complete, I shall maybe be able to do some long-distance tests. It’s rarely made clear, but the RC server versions usually have the code which gets into MC, so the whole grid will have the new region-crossing system. Dare I try full throttle on my F-111?

    Like

    1. Communications: absolutely no argument from me – as you’ll know from reading this blog :).

      MC roll-out: there do appear to have been a high number of caching errors & I’ve had some reports directed to me of problems (communications again – people turning to bloggers rather than the forums…). You’re also not alone in the incidence of controlled movement / flight ocer regions which were not handshaking correctly with neighbours. It was raised in the Simulator group, but viewed as a one-off. I’ll make sure I feedback more detail on such matters in the future.

      The region crossing code may actually make little visible difference to vehicle crossings, other than a faster recovery when rubber-banding or loss of control are concerned. This was certainly the situation those of us encountered during the “pile-on” tests in December 2012. As I commented in my report at the time, the results were disappointing. Avatars on foot appear to do a lot better, particularly those with heavy script loads, and regions didn’t seem to stutter with a lot of too-ing and fro-ing. It’ll be interesting to see the feedback as things settle down through the week.

      Interesting you raise the status of code across the grid post-deployments & MC code; this is something I’ve been meaning to mention for the last few reports which coder the server roll-outs, but keep forgetting to add as a “footnote” to my deployment notes in those reports.

      Like

  2. I still see no clear articles on what the Mesh deformer does, so i’ll wait until its out then i guess ill find out

    Like

    1. Simply put, it’s a means to enable mesh clothing to “better fit” avatar shapes. It won’t necessarily eliminate the need for alpha layers in all cases, but it should help ensure mesh clothing fits more appropriately to a range of avatar shapes, rather than people having to mask their shape in order for them to fit the clothing.

      You can read the original intention with the project here (although the scope has changed)

      There is also an interview with Karl Stiefvater wherein it is discussed, and Karl produced a couple of videos on early development, like this one.

      There’s also some discussion of the “standard sizing” debate here, which might help.

      And more general links and information on the deformer as I’ve reported it gathered under the Parametric Deformer tag on this blog.

      Hope this all helps :).

      Like

Comments are closed.