Mesh development continues moving forward in Second Life. The rumor that Mesh release is planned for the end of May continues too, in spite of the rumor having been traced to a Linden running a contest for building a mesh avatar that ends then.  There are some new developments.

SL Mesh Upload Costs

Mesh Cost – Equivalency

The current Mesh Project Viewer is reporting high prim costs for mesh uploads. Those numbers are buggy. An 800 triangle mesh is not going to cost 3 to 6 prims. That is an anomaly from the viewer and the server being out of sync. There is code in the viewer that bases its calculations on information calculated and sent by the server. If the two sets of equations are not working together one gets bogus information, which is the case right now. So, don’t have a heart attack if you upload a small mesh and it looks like it will cost a fortune.

Also, the upload numbers on cost and prim equivalency are viewer estimates. The actual value after upload is calculated by the server. So, one may see different numbers pre and post upload.

Runitai tells us,

… 2000 triangles in the high lod with a 25% ratio and a 1m size equaling 2 prims IS roughly sculpty equivalen.

Texture Dropping

If you build, you know about dropping a texture on a prim. It is sort of a shortcut for placing a texture on a prim. That feature is going to be available with mesh too, at least according to Runitai Linden.

Mesh objects can have up to 8 textures, if I remember correctly. Which face one is working with is an in development part of the Mesh Project. Selecting faces is rather complex with mesh, from a programming perspective.

However, there will be no scaling of textures on mesh or at least not when mesh is initially released. I may have misunderstood this point.

Mesh Simplification

You probably know that mesh is upload as multiple copies of the mesh. A mesh for each level of LoD (Level Of Detail). Also, there is a mesh uploaded for physics. One can use the same mesh for all 5 slots, which will be prohibitively expensive. One can also make a mesh for each slot. The higher the LoD the simpler the mesh should be with the physics mesh being the simplest. A third option is to allow the viewer upload tool to simplify the base mesh loaded for the most complex level (LoD 1).

There are a bunch of settings for that last option. Getting the viewer to do it well is a problem. With various settings one at least has a chance to get it to work for a simple mesh. The details of how the process works are given in these pages: Geometric Level of Detail, GLoD Definition, GLoD Parameter Defintions, and more definitions.

GLoD was picked by the Lab because it is one of the few decimation processes that can reduce the number of vertices and triangles and preserve the UV Mapping. The preservation is a big deal. When one simplifies their mesh in 3DS or Blender they realize how complex the UV Map issues are because they must manually deal with the problem. They have a great deal of control doing it manually. But, it can be tedious with complex shapes.

Is Weighted Handled?

When the viewer simplifies the mesh there is no weighting data that gets simplified. The nearest vertex in the simplified mesh is used as the weighted point from the complex mesh. This will create some odd affects. But, it is a complex issue and too much to hope for in the initial release.

While UV Maps get interpolated weighted information does not.

If you are seeing the possible problems from this approach see JIRA CTS-525. The Lab folks could use examples of weighted meshes with problems. If you want to keep the mesh private and put a copy in the JIRA, make the upload private. Nyx Linden is making new notes in the wiki for how to do that. Give Nyx a shout if you can’t find the wiki information.

Summary

There are still technical problems but the project is moving forward. The Lindens are making no new noise about possible release dates. I take that to mean some of the current issues have no clear completion date. So, patience… (sound of nails tapping the desk)