Mesh development is continuing. There is almost no discussion of anything new. The focus is on getting mesh working and the import and cost calculations worked out. The Lindens are concentrating on bug fixes and getting this initial roll out working.
You can see in the image a couple of Tiny’s. These are tiny tiny avatars. Both are way cute.
Experiments by Drongle McMahon brought him to the conclusion that for objects under 0.5 meters the physics shape silently converts to a convex hull. Charlar Linden thinks that is what the system does, but will need to confirm with Nyx to be sure.
You can see the TBD (To Be Determined) notes and changed buttons. The Calculate weights and fees is a new one.
Mesh Made Shadows
I had not thought of it, but using mesh for high altitude building creates ground level shadows. Chodai Yoshikawa asked if there would be a distance factor or some setting considering Draw Distance tht would limit the casting of shadows.
Runitai Linden answered that there isn’t. He has tried. But, the affect is more ugly graphics artifacts that are undesirable. The Interest List, things your avatar can see, does not consider the Z-Axis. So, there is no easy way to tell by distance what should and should not cast a shadow.
As to having a shadow flag in objects, a sort of I caste or don’t cast shadows switch, that will apparently require a protocol change. So, we are not going to see such a control in the near future.
New Collision Information
Charlar Linden announced that the Lab got permission to publish some of the Havok Documentation information on collision calculations. See: Best practices for creating efficient collision volumes. This is a downloadable PDF file. It is several pages long. If you want to reduce your Mesh Costs, this is a necessary read.
A current bug can render rigged mesh (clothes) invisible. In general if you come into an area wearing rigged mesh the avatars there can see the mesh. Those avatars that arrive after you, will not be able to see the mesh you are wearing. We saw some of that at the meeting.
Default Mesh Name
You may notice that default mesh names used by the viewer use the name of a previously uploaded mesh. This is a bug in the mesh uploader. It is unlikely to be fixed before roll out. People will just have to be careful with naming and pay attention. Some time post roll out the problem will be fixed.
Sculpty Prim Equivalence
There is a rumor that sculpties will be counted as mesh if they linked to a mesh. This would destroy their 1 prim cost in some cases. Runitai Linden tells us that is the case and it will be the case at initial mesh roll out. However, the Lindens are considering changing that. We’ll have to see what happens.
ARC – Avatar Render Cost
ARC is going to break when mesh rolls out. Break may be too strong. It will NOT be accurate. ARC will not be fixed until sometime after post roll out. So, an avatar wearing mesh will not produce an accurate ARC value.
Mesh Release Date
The Lindens are not giving us a definite date. Charlar Linden said in the last meeting that they are on track for a July release of Mesh… I suppose that could mean just the release channel. I know the Mesh Prep server update continues to improve but see roll backs. I’m sure no one is absolutely sure when Mesh will reach the main grid for general use. Charlar says there will be an annocment about mesh on the main grid very soon.
Mesh Upload Fee
We still do not have final mesh costs. However, today it was mentioned that several people had noticed the upload fee had been cut. It was previously L$150 per mesh. Charlar Linden told the group that the lower upload cost is for the initial roll out. I suppose that means the cost will go back up. Charlar says costs may change after the initial roll out. I would say that is a good bet.
Monday’s Mesh Viewer is version 184.108.40.206005. I checked out what the new upload fees are.
It appears there will be a window during the initial phase of the roll out with cheap upload. I am guessing the Lindens want as much mesh uploaded as possible during the initial roll out. The idea is to see how people will use it and what kind of creative, or dumb, things people will do with mesh. So, if you have mesh stuff ready to upload, get ready.
Also, while I am sure the Lindens will do everything possible to avoid breaking mesh uploaded during this period, I would not assume that it is an absolute. With the upload of all manner of mesh we may find the Lindens over looked something and need to make a correction. So, uploading a couple of billion dollars of mesh may not be a wise idea. On the other hand it may be a smart thing to do.
My Cost Experience
A model I’m playing with costs L$12 to upload. In the image you see the one mesh and one prim object. The physics weight of the mesh part is 32 triangles. The high level LoD is 719 triangles. Without doing much to optimize the model it comes in at about 1.7 prims. The final equivalency is 2 prims.
I need to add a prim for the instrument read out panel so I can run an animation. You see it in the image, the square thing at the top. It needs to be scripted to do stuff for me and control other things. The idea being this is part of a Myst style puzzle and it will power other things. Adding the prim only, my cost remained at 2 prims. Adding the script my equivalency goes to 4 prims. Double checking I can remove the script and reduce my cost by 2 prims.
Building a similar object from prims would take 6 to 8 prims depending on how much like this one I would want to make. My final will take 2 textures. On a prim only model I would need 3 textures to get a close look-a-like. I won’t go into texture sizes and precise comparisons. But, I think I would save tier and get better render speed with the mesh over the prim only model.
Thanks for this post. It helped me get a better idea of how mesh will impact me inworld.
Regarding your experience with the Myst-like puzzle model, I’m curious why you decided to add a prim for the scripted control, rather than adding a face to the model.
Also, just in passing, a Mesh advantage in the steady-state rendering speed may be offset by the initial model download cost. One can push a lot of parametric prim values down the pipe in the time it takes to send disappointingly few vertices.
This is no big deal for one Mesh model, but scenes will accrete models (just as they did sculpties). If the PE numbers reflect the lower steady-state rendering costs, rather than the higher bandwidth demands, we’ll soon find ourselves looking at a gray, blobby world (again).
Much thanks for relaying the info about high-level shadows. That explanation is pretty discouraging. (BTW, resurrecting the long-deprecated PRIM_CAST_SHADOWS parameter would not be an ideal solution here; one would still want to see shadows cast from high-altitude objects, when viewed at those high altitudes.)
For your curiosity… Using an additional face in the mesh is simply beyond what I know how to do right now. I know it is easy. I just haven’t taken the time to learn. In this small model it does not seem to have a prim cost impact. But, there are some combinations using another face I have not tested.
I agree that prim parameters verses mesh vertices download may give prim some advantage. I suspect that is why the Lindens made a change in the cache and are working to improve caching. That may reduce any delay with mesh download. But, even large (many vertices) meshes are not as large as most textures. I expect the download time for mesh to be small. I’m not sure how much over head is involved in each request. I suspect the additional requests may produce more asset server lag than the actual download. We’ll see.
Pingback: Ode to Mesh… Fit the Fourth « soror Nishi