Second Life – Third-Party Viewer Developers UG 2017 w/20

Oz Linden has his 64-bit viewer crash for the first time since he’s been using it. I’ve used it for weeks and only seen a couple of crashes.



The HTTPAssets Viewer got an update in week #20.

Voice is the same version. It has an ‘elevated’ crash rate. They haven’t yet figured out why. So, it won’t be promoting any time soon.

The maintenance viewer has the new Estate Manager controls. This is the more likely viewer to get promoted.

Expect a 64 bit RC release in a couple of weeks (#22 or #23). The last functional update has gone to QA. Expect it to update in week 21.

This version of the 64-bit has the new installer. It will check the computer and viewer versions and install the correct version. If no install or update is needed it will launch the viewer. The installer then just waits for the viewer to exit then shuts down. Oz is planning to move much of the logging and crash reporting out into this installer process.

Consider that this will give an improved opportunity for crash reporting. When crash reporting is part of the viewer and the viewer crashes it often takes down the crash reporting process. A separate program would avoid that problem.

The number of people using the 64 is only a few hundred. That is because the only difference the 64-bit offers is significantly fewer crashes. The stats on Win 8 & 10 conflate,  but the combined rate is half the 32-bit rate. Mac’s show little difference.

360 work has resumed. The Lindens have added one new library for 360 meta-data. Project update coming soon, couple of weeks then an RC quickly after. The coming changes will make saving 360-images much easier. The meta data allows sites that display 360-images know the image is a 360-image.

During Fantasy Faire there were a lot of region crossing hand off problems… It was so bad I gave up walking across region boundaries. I TP’d from region to region. The Lindens learned a lot about what was going on during the faire. It was not a new problem. They had been looking at the wrong things but found the thing to look at.

The server logging has been updated to watch the problem, which is the Caps Router… Previously they could only get error data after the Caps Router had run out of routes (connections) to hand out. Watching this information will allow them to figure out how many connections they need. This appears to be one of old time bottlenecks. It looks to be easy to repair. Doing so my raise the population density a region can carry.

The limit running out is the HOST’s limit not the region’s. The Lab hosts a number of regions in a server/host. So, if you are on a deserted region that is in a host where other regions in it are crowded, you could run into the hand-off problem trying to cross into a different region.

Next week’s RC rolls will have some of the correcting changes. Making these types of changes that removes an old road block can allow a flood of traffic to rush down the road to a sometimes-narrower bottleneck, creating a new problem or problems. So, the change process is done slowly and carefully.

Another problem fix going out in RC’s is one they noticed in the Fantasy Fair issues. When a teleport is rejected by the destination region it does a bunch of work. To reduce that load, they will first change the throttles on teleports. That may break the Teleport HUD’s that hammer their way into a busy region. Or at least slow them down.

The new population limits that allow premium members into a full region ahead of others are now grid wide. So, I’ve been enjoying my exalted 😛 status and visiting more events without waiting. The TP-attempt-throttle will further degrade the usefulness of TP HUD’s.

Fantasy Faire 2017 - Fairelands Junction

Fantasy Faire 2017 – Fairelands Junction

The idea now is to reduce the work that regions have to do to keep people out. That will allow those in the region to have a better experience. It also adds the likelihood that a region can handle more population.

Interesting trivia: the operating metrics graphs for the SL system display a real-time TP Fail Rate.

There was some discussion about what causes TP failure. Inventory size does not affect TP. Wearing a lot of stuff does. You could get a time out fail when wearing lots of scripts and huge meshes.

The Lindens looking at ACI are making progress on reevaluating the render costs. Data is incomplete. They are measuring a number of things and their effect on render speed, one of which is the number of polygons worn. So, I expect ACI values will change at some point.

Our viewers report our ACI setting once per minute and reports the avatars it is rendering.

More trivia: Oz has been with the Lab for 7 years this week… wow!

Leave a Reply

Your email address will not be published. Required fields are marked *