Earlier quoted context omitted.
What you describe exists in opensimulator worlds. It's friday night, so a lot of virtual parties going on: https://opensimworld.com/events
Cool, this is closer to what I think will ultimately be successful. From the FAQ: What is OpenSimWorld? OpenSimWorld is a directory of 3D worlds that use the open source OpenSimulator software and are connected to each other via the hypergrid functionality. OpenSimWorld is not a Grid, and we don't host or rent regions, we only provide an easy way to discover amazing virtual worlds and to advertise your own region.
A big part of the problem with attempts to create a metaverse, and the reason Facebook's efforts look so crude, is that it's very hard for ordinary people to create high quality and performant 3D content. Getting solid framerates in AAA games requires the art and tech teams to work very closely together, and for art to be constantly measured against performance goals in every scene the player could view. Also most AAA games are set indoors to limit draw distances. This kind of hyper-coordinated and planned effort just isn't compatible with a free-wheeling metaverse, so SL constantly had problems with performance.
For example, most people imagine the metaverse as starting from an open Earth-like world, therefore it is largely outdoor spaces. This implies potentially huge draw distances especially if you want a geographically contiguous space like SL had. Additionally, people loved to create objects that had way too many polygons, or textures that were too large, or more subtle problems like buildings with translucent windows. All this stuff seems fine and simple if you're a regular end user but to 3D rendering experts it's just a disaster because it requires so much overdraw.
Another key problem SL had was that the entire world was mutable, because the act of playing the game was the same thing as the act of building the game. AAA games rely heavily on large batch computations that optimize essentially static worlds with very carefully controlled mutation layered on top. For instance static lights are much cheaper than dynamic lights because you can 'bake' static lighting into textures. SL/metaverse isn't compatible with this sort of optimization process because at any moment someone might decide they like the light a different color, or in a different place, or to script it.
Finally, you have the tension between openness and professionalism. The utopian metaverse vision assumes a kind of roughly equal ability to create things, not just a small elite class of pro builders. The problem is that the better you make the world building tools, the quicker your population splits into a tiny creator class and a much larger passive consumer class. For instance SL had (has?) a small economy of people make and sell objects, in particular "clothes". Although in theory anyone can make anything, the more sophisticated your metaverse becomes the more quickly it becomes a consumer oriented culture. For a social network like Facebook this is problematic - the whole reason they got so huge is because social networks are great levellers. Twitter is of course the epitome of this approach in which you are literally prevented by software from writing anything long enough to be truly outstanding, but Facebook's design was very similar. There is no such thing as a professional tweeter or professional Facebook posters (the much more image-oriented Instagram does have professional posters).
So the ultra-crude Mii style approach is probably not the result of stupidity or lack of vision but rather an attempt to learn from prior metaverse efforts:
- Low quality graphics make it easy for people to create content.
- It sets expectations; this is not a world trying to look AAA quality and failing, it's not even trying.
- It frees up rendering resources to chew through extremely unoptimized content at a good frame rate. Especially vital if you're trying to achieve VR framerates!