Viewers and Mesh

The opinions on Collada mesh coming to SL are mixed. Hamlet put a set of links in a recent blog post calling them ‘gems’. I find some are and some aren’t gems. But, digging through the chaff usually reveals nuggets of neat information. Here are the current tidbits.

Mesh DOA?

One line of thinking is that Linden Lab has again botched a new feature. Some people would be unhappy if the Lab sent them chocolate, cherry filled orgasms. When one starts from the premise that the Lab can do nothing right, it is hard to see what is being done right. These negative people are leading me to the conclusion that at an SL age of 3 or 4 one becomes an old curmudgeon.

Whatever the case, there are some interesting tidbits of information leaking out. 

In the comments on one blog Tonya Souther, of the Phoenix team said… (link)

There are no doubt other TPV teams looking at adding mesh to the 1.x codebase. I know we on the Phoenix team are not; we’re concentrating on making the current viewer stable so we can concentrate on putting out a v2 viewer with a UI that doesn’t suck. We’re just beginning to look at that now, since we’ve spent the last 6 weeks just trying to get something good and stable for people to use while we go off and spend 6 months on v2.

I guess this means we will see a pause in Phoenix viewer releases. Getting setup and forging a new team to develop the Phoenix viewer has been no easy task. The new 373 Phoenix is much more stable. But, this statement leads me to think it will be some time before we see mesh support in the Phoenix viewer.

It will be some time before the Lab releases mesh to the main grid. So, this may not be a problem. However, I can use mesh on the OpenSim grids now. OpenSim users are looking for a viewer that will handle mesh now. I’m using the SL Project Viewer (SLPV) on OSGrid now and experimenting with mesh.

Also in comments Ron Overdrive said…

The part about TPV’s not being able to upload Mesh isn’t 100% accurate. Yes we won’t be able to use their Convex Decomposition library which uses Havok to map the physics on meshes, however, they’ve provided a reference stub in hopes that OS devs can find a solution they won’t dedicate their time & money to themselves. There’s already devs talking about adapting Bullet to their Convex Decomposition stub to generate the physics details for meshes. So Mesh in TPV’s will be possible, just not immediately.

This comment is in response to a licensing problem LL has with part of the mesh upload software. The Physics Decomposition part is reliant on Havok physics software. How this is going to be worked out for TPV’s is a problem yet to be solved. It is much like the KDU problem. Work-arounds were found. So, history suggests this is not that much of a problem. Some developers are already working with Bullet, a free open source physics engine, to make an open source replacement decomposer. Those ripping on LL are into the idea the Lab is trying to screw over the TPV Dev’s.

Physics Decomposition is a process that looks at a mesh you are importing and tries to figure out how to handle the physics. This is the physics that allows you to walk over or bump into the object made from the mesh. You may build a complex table with odd shapes. Sending that shape to the physics engine for collision detection could bring a SIM to its knees. So, the importer decomposes the shape into something simple, like a cube or a couple of cubes, a cylinder, or something. But, you also have the option to design your own simplified physics shape for your mesh. You do not have to rely on the built in decomposer. You will likely get a better result, if you do. So, a built-in decomposer is not absolutely required.

Shadows

The SLPV-Mesh has working shadows, if you use nVidia. Also, anti-aliasing (AA) is working in it. The current SL Development Viewer (SLDV2.4.0-212775) also has AA working. Oz Linden says it is now in the 2.2 viewer pipeline. I think that means it’s in the 2.2 Beta version. However, there is a problem for Mac’s that has not been resolved. Q Linden has found using AA and Full Screen is not working. The plan is to have this resolved for all platforms by version 2.3.

Chat Logs

They are a problem. They are changing over to LLSD format. See: Second Life Chat/IM Log Format. This change apparently came out of one of the teams not yet using the new ‘open and transparent’ process. The change was put in place before the support for it was in place. There seems to be some of this going around. I guess old habits die hard.

Double Click a Panel Border

Odd. Just odd. Apparently double clicking on the new panel borders lines up the panels. I would never have figured that out.

Light, Shadows, and Shine

Seems something has changed in how lights and shine work in the project viewer. I think this is a buggy part of the viewer. So, how it will shake down one can only guess. But, you will see very different results depending on whether you have deferred render on or off.

Shiny has been a kind of faked reflection. The new viewers are supposed to render shiny as real specular highlights. This means the white spots on black latex outfits would move as the AV moves. At some camera angles they would disappear.

Many people have asked for real reflections. For a time there were reflections, back about viewer 1.13. Windlight changed the game for reasons I don’t understand.

I doubt we will have general reflections in SL because of the computer power required. But, we may get mirrors some day.

Lights are buggy right now. See KirstenLee Viewer S20(40) Released for an example of the current lighting bug.

Mono Scripting Freeze

You probably know that rezzing Mono scripted objects can freeze the SIM for 5 to 30 seconds. Lab staff has been chasing this problem since it was encountered last year. The recent changes to the server software and the TMalloc changes is allowing more understandable debugging and tracing. It is looking like we may have a fix in the next few weeks.

Kelly Linden writes that SL is running on Mono 1.2.6. They are in the process of testing Mono 2.6.7. This is supposed to come out in two stages. If I understood, 2.6.7 will be one stage and then moving to the current release as a second step. Kelly thinks they are on the trail of the last problem in the System.InvalidCastException errors. If they can nail that one, we will see the change coming to a test region near you.

Viewer Code for Meshes

The Lab ran into some licensing issues. So, not all the mesh code made it out on time. Nyx Linden writes it is on its way.  The Physics Decomposition code mentioned earlier is a concern of Nyx’s. It is a proprietary section of code they cannot release. An open source replacement will have to be devised. Those working with getting a TPV Decomposer should contact Nyx for help.