Prims to Mesh

Mention has been made of possibly getting to import meshes at some point

Might we be able to convert objects made from prims into meshes and perform hidden face removal and whatever other optimizations might be made to the meshes and then use the optimized mesh versions?

Would it be possible to do a "Convert to Optimized Mesh" in place in the viewer?

This is possibly one of the features we need the most. Sadly it would take a lot to implement that into Second Life I think.

One: Second Life needs to properly support 3D mesh objects. This includes uploading and storage of information, proper restrictions so overly detailed meshes cannot be uploaded(control lag), Proper LOD control mesh object side, Probably more but this is what comes to mind late at night

Two: Second Life will need a complete system to analyze a 3D object, then perform boolean operations, merge and create texture maps, then spit it out in a simple format for the user.

Three: if all that was implemented, i would think the next step would be to bake shadows into a scene!

guess the overall thing that needs to happen is laying the foundation.

The concept of using meshes in SL has impacts on so many levels in SL, that it's not something that will see pretty soon.

The technical impact.
Meshes cause a complete rewrite of code on multiple levels of the SL architecture.
On the network level, SL uses (used) a fixed packet size model to send prim/object data to the client, the whole initial idea of SL was to create a 3d environment which could be 'packet streamed' to clients. Meshes will not fit in a 'single fixed packet' model, so this needs to be rewritten to use a multiple packet system and recombining packets on client level. (A process that's underway anyways, with the switch from UDP to TCP/HTTP standards)

On the DB/Asset level, this requires a lot of extra code to prevent 'misuse'. Where in normal 'gaming' environments vector optimizations are a constant focus by designers and developers, in SL, this has to be constantly taken care of by the architecture itself. Adding meshes to this mix will make this process a lot more complex. (Which will cause a lot of extra processing cycles spent on this too...).

The LOD issue. It's pretty complex to create a reliable LOD mechanism for arbitrary shapes, we're already seeing major issues with this on sculpties.

The social impact:
Firstly, with sculpties we've already seen a 'split' in the community between 'those that can' and 'those that can't'. The skills to create (and especially texture) meshes is only available to a small part of the community. SL has changed from a place where 'everybody can create' to a place where 'everybody can dabble, but only a select group can really create stuff'. Over the years, SL has transformed from a place where everybody was building, to a 'consumer/producer' environment. Meshes will make 'prim building' something of a lesser skill, immediately devaluating a lot of content produced in the past. Of course, this is 'progress', we can't mourn the fact there are no skilled 'carriage builders' anymore either. But as a 'prim content' creator for many years myself, it still feels like kind of a loss.

It also creates a situation where content creators don't have to spend any time in SL itself anymore, all aspects of creating, except for the final 'upload' are done outside of SL.
People constantly complain about how LL (the creator) has lost touch with the actual residents. By moving the content creation process more and more out of SL, the same can happen with the content creation community.
I'm split about the whole idea of moving more and more of the content creation process out of SL and into external tools. It's good from a quality standpoint, but it makes for a world where we're just using the stuff of a large group of creators that have no real tie to SL and it's communities itself anymore.

It's possible but it's far from easy because of issues relating to UV mapping and LODs (reducing detail at a distance).

So is it worth it?:

Many of the primitive models that would most benefit from being converted to meshes aren't really worth saving because they're fugly. They're still going look like a bunch of primitives stuck together.

Architecture (prefab houses, tables, sky screens and sex balls) wouldn't really see any benefits from being converted to a mesh because there aren't many hidden faces.

Also, converting a prim to mesh would only be useful if the textures on them are 'baked' into a new texture for the whole mesh. The resulting texture would probably be much larger on most prim structures as the individual textures used on the prim structure. (Except for untextured/flat textured items.) So there is no benifit from a performance or bandwidth perspective at all. (The creation of an optimized UVmap/texture for one or multiple mesh objects for games is currently still mostly a human skill.)

The conversion would also make any 'mod' item useless. As modification would either mean a complete redo the baked texture or a complete rebuild the object.

As others have stated, the important question really is not whether it would be technically possible to convert a prim model to a more efficient single mesh, but whether there'd be anything to be gained by doing it. No matter how slick or easy the conversion process might be, there's really no getting past the fact that it would never be worth doing. I'll explain why.

First of all, if one wants to build a mesh model, SL's lego-style building process would be one of the worst possible ways to go about it. Compared with conventional poly-modeling techniques, prim-based construction is unnecessarily time-consuming, very inefficient, and fraught with inherent difficulties that simply don't exist elsewhere. It adds an unbelievable amount of complication to what would otherwise be routine tasks, like UV mapping.

I realize the motivation behind the question probably had more to do with concern for pre-existing models than for new ones, and while that's certainly a valid sentiment, the only really good response, cold as it may sound, is just to say let them be. One by one, they'll either get replaced by much better mesh models, or they'll stay as they are, to become relics. There's not much point in trying to convert them, since the results would never ever be anywhere near as good as brand new meshes made from scratch, on any number of levels. There's absolutely no way around the fact that a mesh designed to be a mesh from the ground up will always be far superior to one that was converted from any other type of model, with prim-based models being among the worst of the worst. That won't change, no matter how sophisticated a converter might come along.

Once meshes hit the grid (which likely won't be for a very long time yet), there simply won't be much point in prim building anymore.

It would be interesting, but there would probably be some kind of "limit" to how complex it could be.

With people complaining something simple as SL can barely run on their computer, I can just imagine the complaints when they run into a 350000 face mesh.

Or- they could go the "There" route, and you'd have to get LL to "OK" your mesh before its allowed on the grid.

You could expect ALOT more people taking the easy way out, downloading free 3D models & just uploading them.