How to avoid LOD mixups in multi part mesh uploads ?

[Mikki Miles]

yes, I learned this on my journey to where I am now, though I could not name it that clearly like you did ;-)

it's the step to export LoDs for DAE-files with independent object's, which gives me headaches and crashes my viewer. But, thanks to you and Drongle, there are some hints to watch, sooner or later I will be there ;-)

Hopefully the exporter in Blender 2.5 will be able to export proper files for SL-upload in the future... 

[Drongle McMahon]

I think I worked out the crash condition.  If you have an slm file and with meshImportUseSLM true, then if you enter a file into a LOD slot with the order of the geometry tags changed, you get an immediate crash. This can happen (a) when you first use the sorting patch and there is already an slm, or (b) when you change the name of an object so that the sort order changes. I tested it with manual file edits, just changing the name and with just changing the order, and it was the order that caused the crash, not the name change.  I made a jira (or add to an existing one).I will also add advice to the patch post.

[arton Rotaru]

Awesome! That works just perfectly for me!:matte-motes-sunglasses-3: Only little drawback is simplifying of the physics shapes. It's a bit harder to achieve the simplest decomposition which is still as accurat in shape as a single mesh. But that's only small differences in PW.

[Drongle McMahon]

Yes. If it were not for the effect om upload fees, I would stick to uploading multiple objects individually, to have maximum control. The sensible reduction in the per-upload-session cost (now L$10 instead of L$150) makes that difference smaller, but there is still a significant saving in uploading models together. Of course if it's going to be a linkset, then it saves you a lot of assembly work too.

One thing does concern me though. Things like my gallery use 2-4 copies of some component meshes. Uploadingt these separately, each is uploaded once and the replicated in-world. I would probably save upload fees and work by uploading the complete thing, and replication of these meshes in the upload would add little to the cost and save much work, However, the replicates would then presumably be different assets instead of instances of the same one. This would cost nothing but would consume resources because the different assets for identical meshes would all have to be downloaded.

I think it would be good for network performance if some sort of incentive for instancing mesh, preferably with all three simple (scale, rotation, position) transformations, can eventually be developed..

[Gaia Clary]

if LL would supply an upload descriptor (idealy XML based), then we could create an upload set which takes care of instances (avoid duplicate uploads), and either create an LSL file (like we did with sculpted prims in the past) or create an assembly list based on LLSD (as we will support in the future for all prims). See also CTS-634

But i still prefer to see such caretaking of duplicates done by the importer itself.