I would like to request the ability to layer materials on the base avatar. From a user interface perspective, this would show up as a list of layers. The one on top covers all layers below, and so on down the list. The user could drag and drop a layer to a different place in the list.
From a creator perspective, this would entail importing what I'll call a "sleeve" representing a tattoo, painted-on garment, or even an entire skin. The sleeve would start out as a mesh template the designer downloads. They would apply materials to it in MD, Blender, or other programs, but without modifying the mesh at all. One special map they would create is an alpha map showing opacity. If they wanted to create a T-shirt, for example, the alpha map would indicate that all the body that is not T-shirt is transparent and vice-versa for the T-shirt part. During import, Sansar would strip away the mesh and extract all the textures used for alpha and materials or let the user manually select hand-crafted textures for them. I imagine that Sansar would need to resample the textures to suit its own pixel density requirements.
The avatar can be "baked" in the SL sense to keep avatar rendering performance high, but I think it would probably suffice to just let every client do the baking once when the avatar appears. That way we can avoid the debacle SL introduces of delayed and sometimes incomplete avatar baking.
I think it would probably help if there was one more special alpha mask that could be called a "hider", maybe, or a "holes" layer. The purpose of this would be to carve away all layers of the avatar, presumably to deal with meshes that would benefit. Imagine an attachment of a prosthetic leg, for example, that needs to be narrower than the original leg. And, of course, the more dramatic example of completely blanking out the original model to replace it with, say, a furry mesh avatar. That isn't the ideal way to do a furry, IMO, but it would be one way.
Okay, so this is a rough proposal from someone who is not a fashion designer here. I'm putting it here to solicit input, especially from those of you who are. It seems to me that the alternative being employed now of creating a mesh that is slightly thicker than the underlying avatar in order to cover up that avatar is wackier than just allowing us to reskin the avatar. And this mechanism I have described has a huge benefit of allowing us to combine layers, just like we can combine attachments now. And with no apparent hit to runtime performance, once the layers are baked together into final textures.
What do you all think?