Second Life News 2013-30 #2

Some of the most interesting news comes from the Third Party Developers meetings. This is not a meeting that is well publicized. I suspect many think that having just the well informed attend is more productive. I tend to agree.

This article is 2,600+ words. It has some work-around methods for some current problems that people are experiencing. It covers some of the most informative news about SL that is available.

TPV Dev Meetin 2013-30

TPV Dev Meetin 2013-30

Server Side Appearance

SSA is going extraordinary well. The stats coming in are as expected, which means the backend is performing as expected. This means the Lab’s estimates for the amount of hardware needed are going to be correct.

View adoption is going pretty well. About 75% of users have adopted an SSA capable viewer. (84% of Firestorm users – see chart) Looking at that the other way, something like a quarter million people login each day with an old viewer. Also, the numbers vary quite a bit from day to day. 

Still the adoption rate is surprisingly high, at least it is surprising me. But, there is a lot of effort going on to get people to update. Having the Linden splash page at login carry a big warning is probably helping. I haven’t looked lately to see if the Firestorm Viewer’s splash page is carrying a warning. I suspect it is. So, between those two viewers the majority of the users are exposed to the information.

As of Friday the 26th, the Lindens doubted they would more SSA to the full grid. By Friday the server software candidate is pretty well decided on. Data collected over the weekend drives the final decision made about mid-day Monday. We see that decision published late Monday in the Server section of the forum as the Deploys for the week of

So there is a good chance things can change until late Monday. But, it looks now as if the SSA server update will remain in the RC channel for 1 to 2 more weeks.

Bug

SUN-98 – Bake Fail Resulting from Partially Broken Alpha Layer. – As far as I know this is the first known Bake-Fail in the new SSA system. It may sound odd, but the Lindens are considering this an ‘expected behavior’ for now.

What is happening is: a corrupted Alpha Layer, the asset in the system, being corrupted gets skipped by the SSA oven and is not baked into the composite appearance texture. This is expected behavior because the Lindens do not want corrupted textures used for appearance. The result is unpredictable.

The fault is not actually SSA’s. It lays in the asset system. So, the asset system will be tweaked for a fix. The behavior of SSA with a bad texture is as expected. BUT, you could still end up with nude or partially nude avatar, unless it is the skin texture that is corrupt. I’m not sure what happens then.

BUT…but…

You may remember that SSA is supposed to stop bake fail and accidentally nude avatars. Well, that is part of the reason for building SSA. From the Linden side of things they have accomplished solving the problem of accidental nudity from bake fail IN THE BAKE PROCESS.

One cannot realistically expect to feed garbage into the bake system and have it turn out a wonderful product, garbage in, garbage out.

So, this is another case of different problems causing the same symptoms, this case accidental nudity.

The Solution

Still the problem needs to be fixed. Consider how you fix a corrupted upload that has been setting there for maybe years. The pre-SSA viewers were sort of handling the problems and giving us bakes of some kind. The newer more aware process is giving us better bakes but seems to be deliberately less fault tolerant.

Nyx Linden expects to add error reporting to the system in the next phase of updates, whenever that may be. I expect ‘real soon.’ So, the SSA system should do something like send you a notice as to why you have an appearance problem.

WARNING!

Your boobs are hanging out!
Texture on shirt [item_name] could not be baked
due to file corruption. Sorry.
Nice set of Tangos.

OK… that probably won’t be what the Lindens actually come up with for an error message. That is a bit Mayor Filner-esque. But, there will be some message to give users an idea of why there appearance is failing to properly bake.

Bug

There is SUN-99 – Bakefail only on SSA regions. When entering SSA region, skin, and system clothes fail to bake. – This is a case where no matter what one does their appearance is failing to bake or their appearance is stuck. The main cause is having multiple Current Outfit Folders (COF). That could happen early in the life of COF when some viewers would allow you to move the COF and then the system recreated the system folder.

Some people moved the COF into their clothing folder, which seems reasonable. They are only now seeing the gotcha associated with that change. Now that more restrictions have been placed on the folder, they cannot delete it or move it back to the default location. What a mess.

You may find you look OK to yourself. However, others may see you differently.

So, if you are seeing or people are telling you of appearance problems with your avatar in SSA regions, check to see if you have multiple COF’s. Then contact SL support for a fix.

For now the Lindens are looking to see if they can automatically fix the problem via SQL in the inventory system. Fortunately it is a rare occurance.

Also, the Lindens are looking at ways to build in handling for multiple COF’s. Mostly to handle correcting the problem.

SSA Viewer Changes

Nyx Linden has said there are some server side and viewer tweaks being made but nothing significant yet. These changes are being described as polish. This means that we will likely see one more set of SSA viewer changes roll out. Since these are a priority updates they will likely crowd ahead of other Release Candidates.

There is some fuzziness in what the Lindens are saying. If there are any problems found in SSA, then there is likely to be more code coming and that likely means two viewer releases due to SSA. But, that is probably the Lindens double plating their CYA layers… I can hope.

For third party viewer dev’s its unclear to me whether this is good or not. It is going to give them more time to get their viewers tested and polished. It also means they don’t have all the code yet. 

Removing Bake Code

We are hearing that the Lindens will be removing viewer-side appearance baking code. To some extent that is true. But, not all the bake code comes out. Once SSA rolls to the entire grid and Lindens are satisfied it is working, much of the viewer’s baking code will be removed and cleaned up.

However, when you first change clothes the appearance you see was and is baked by the viewer. When you change clothes there is a chance for a double blurry. In my experience the second blurry is brief and in some cases has not been noticeable. This second blur is from downloading the SSA baked texture and rendering it.

The code that makes the initial bake is staying in the viewer, AFAIK. The code that is staying reduces the load on SSA and you see your appearance faster when changing clothes. At some point a signal is sent to the SSA server and your appearance is baked and sent to you and anyone that can see you.

The error notice I mention above is based on the idea that the viewer must request the textures from the asset servers to bake your initial version of the appearance. It is at this point that it should be possible to create an error message for bake problems.

Also, third party viewers will need to retain the full bake code to work with other grids, which are unlikely to implement SSA. They may surprise me. But, adding another server side set of hardware seems costly and thus unlikely.

Vivox Upgrade

Release Candidate is out and the code is now public. I think the release notes are missing so far. So, we don’t know what this fixes. No word on how good or bad the new update is. Some says the changes are minor. No user visible changes. It is supposed to be more reliable.

Firestorm support is dealing with a lot of problems with voice over week 30. They are hoping this new version will be more stable.

If you are testing the Vivox RC, report errors in the general BUC JIRA.

Interest List

Oz expects an RC with these changes to appear soon. Andrew Linden has been absent for a couple of meetings. One was at least vacation. The last is likely due to his working on final touches to the Interest List.

My previous article (Second Life’s Annoyances) points out some of the texture render issues people are experiencing. The Hope is getting the viewers and servers in sync will solve some of the current render issues.

Firestorm

There next Firestorm release is dependent on what the Lindens release. Jessica says they are making good progress on getting the CHUI code integrated. Depending on when the SSA ‘polish code’ is released the development path for Firestorm will change. So, if SSA takes a bit more time there may be a FS release with the new CHUI code. If not they may push SSA out in the next release and CHUI in a release after.

If the SSA takes just a bit longer and FS can roll CHUI, the FS viewer could catch up with the Linden viewer. But, a ‘bit longer’ sounds like 2 months according to Jessica. She thinks they are close to putting the FS CHUI version into their alpha testing soon, one or two weeks.

However, the Interest List code is in hot demand. As I mention above it has the possibility to solve some render problems. Jessica believes that when the Interest List code is released in the SL Viewer the pressure will be on the FS Dev Team to get a release out for FS. She is probably right.

The FS team thinks once they get CHUI integrated they can keep up with Linden releases. Of course that is caveated by saying, ‘Provided there are no other hidden projects waiting to spring out.’

Oz admits there are hidden projects on going. BUT… doesn’t see any that have the wide ranging impact that CHUI had.

Coco Viewer

This new code is coming out in an RC ‘shortly.’

CHUI

Chat Hub User Interface is getting new code. It will be in a new RC soon. I have no idea what that may do the Firestorm timelines.

Materials

The Lindens are working on improvements and fixes for the Materials System. But Oz Lindens feels the next RC for Materials is a ways off.

Viewer RC Pipeline

Because of all the changes and fixes being held up by a narrow pipe for viewer releases lots of stuff has stacked up. Now that the pipeline is more of a river than a pipeline, it can spread out into the delta. So, lots of stuff is going to show up.

For a time there will be lots of simultaneous things happening. Eventually as the backlog washes through the number of things going on will get down something reasonable. 

SL Stats

Internally the Lindens are changing over to a new stats system, which the Lindens say is hugely better. Once that change over is made the TPV Dev’s will get better hardware stats. Like how many people have ALM enabled.

This system was already supposed to be implemented. But, events have delayed it.

HTTP

Monty Linden has been working on mesh upload and downloading. The new code will use a quarter of the connections currently used. So, the current 32 connections the viewer is using will drop to about 8. A thing called ‘keep alive’ saves a connection that has completed the download it was established for and  uses it for the next download. This should allow more older routers to work with SL without problems. So, router restarts could be a thing of the past. This should also allow better connections in large local networks like colleges.

Monty says there will need to be some server side changes and those are currently behind SSA changes.

There are also viewer side changes. Monty is hoping to have those out ahead of the server side changes. Still it will be late August or possibly September before those are out… which seems to also put a timeline on the server side releases.

After this round of HTTP improvements Monty will be adding HTTP pipelining.

There are currently some HTTP problems. The Firestorm team is seeing some people that repeatedly have inventory loading problems. The general fix is to delete the inventory cache, move to a deserted region, turn down bandwidth (which implies a UDP problem) and draw distance to diminish bandwidth demand, and allow inventory to load. But some still have persistent problems with inventory whether they use FS or the Linden viewer.

Some people are getting relief by using Singularity, which does not use the core HTTP library that the SL Viewer and FS use. But, the HTTP library Singularity does use is known to have other problems. You’ll have to try Singularity to see if it works better for you. It may get you around the problems you see now and you might not run into the problems it can have.

The FS team will be collecting logs from people with the problems and filing them in a JIRA. Monty will be looking at them and trying to figure out what is happening.

There is a timing concern. Monty needs the logs within 12 to 24 hours of the fail. I suppose that has something to do with the network logs on the Linden side scrolling the problems out of the logs. To get both sides of the error Monty needs the logs as quickly as possible.

Fix

Whirly Fizz says there is a sure to work work-around for the inventory problem. Open two inventory windows. Drag and drop something from one window into an empty folder in the other window and then relog. I think you also should relog into a deserted region, but Whirky didn’t say that.

After this fix, the inventory problems may reappear with time. But, it is a fix for a time.

Experience Tools

The Experience Tools project remains and invisible project, meaning the Lindens are not talking about it. People like Oz, Nyx, and Monty are not on that project team so they can’t even make generic hints about it like pretty soon or down the road aways. There is currently no information on ETA or scope of changes. Oz daos say it is ‘very actively being worked on.’

We are assuming this is pretty much the project that went into major re-write after Griefer Monday (June 2012). I covered the initial roll out in May in: #SL News Week 22 (2012). The original Linden announcement with a video is here: Advanced Experience Tools Preview. This all went dark after Griefer Monday, other than to say permissions were getting a re-write.

Last week (29) we did get a series of false restart notices across the grid when Linden Realms (LR) was updated. I assumed this update was to get LR on the new Experience Tools server code for testing. But, we have no definitive word from the Lab.

Server Crash Rate

Some information came out that the current server crash rates are the second lowest ever for week 30.

Ed Merryman, a member of the FS Dev Team runs a damage enabled sim. He says that reduces griefing. I suppose that makes it easy to shoot griefer and eject them from the region. Whatever…

Summary

Lots going on. A number of current problems have fixes on the way, There are some work-around methods for now.

2 thoughts on “Second Life News 2013-30 #2

  1. The CHUI project viewer changes are minor cosmetic changes from a code point of view so should not impact the FS team greatly.
    The User Experiences project viewer I downloaded (when it recently escaped from captivity briefly) was interesting, but was clearly a few weeks away from go-live – it will require quite a few changes by the TPV developers.

Leave a Reply

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