I'm not providing a minimap, because I don't really care whether it gets played- I'm assuming it gets played about zero times, except on my rig, for specific testing purposes. And I could give a hoot about comments about the aesthetics, tbh- there were a number of failed experiments in this map (like, for example, "how far towards vertical will texture splattering actually work?" and "what will happen to tesselation when I introduce some noise?", etc., etc., that were in the map purely because I wanted to see what would happen, frankly.
I'm glad everybody cared enough to try it, but I have no intention of turning this experimental design into a full-blown map, because the file format did not do all the things I would want it to do, and after reading what it would take to place Features, I'm not really interested in making a version with S3O objects to make the buildings look nice, frankly- it looks way too much like work for me.
I really wish Hugh's stab at a visual editor worked better than it does, and that certain aspects of the map format were more functional than they are.
Based on most people's results, if your hardware sucks, then instead of getting textures at their lowest mip-levels and no bumpmaps (which is what
should be done, in my opinion) you either get practically no textures and a very ugly experience, or it's broken, plain and simple. Hopefully Jelmer will agree that seeing textures at lower resolutions is a better alternative than not rendering them at all- that would also make a real point to using DDS... whereas, in my experiments prior to release, DDS just meant that even with everything maxed, my textures were blurry, mipped-out messes past some sort of hard-coded limit built into the POV, which really bugged me.
Surely, this isn't how things should work. If this is the map format of the future, then it needs to handle mipmaps more gracefully (scaling with distance, controlled by user settings), tesselation issues need to be handled more gracefully (i.e., at max, you should see something like 16X more polygons and a lot more smoothing, whereas at min you see a very blocky world, but it works)... and I really, honestly think that SM3 should not support terrain deformation on the fly, because that's one feature that slows down render code and pathfinding to a huge degree, based on all tests I've ever performed... more and more modern mods are taking it out of their game designs... and it'd make keeping the mesh beautiful and stable a much more sensible proposition... and I strongly suspect that the cost savings in code speed would allow far more layers to run for people.
Lastly... one major feature I feel is missing, and I alluded to it in my post about the perils of working with this format... is that for me to really take this format seriously, I'd want to see three things that aren't even available:
1. Forboding's uber-giant texturemap of dewm. Go ahead and let us have a texture that is as large as the heightmap or even 2X or 4X or whatever! For a 1024 map, a 4X skin, which would be almost photoreal up to extreme zoom levels, would still run just fine on most modern hardware, JC... and if it doesn't, then the user should lower the mipmap level and still be just fine, because they're not exceeding their texture space.
While I hate to admit that I agree with Forb about anything (
if for no other reason than it's fun, frankly), I will go right ahead and say that, compared to my experience hand-painting NanoArena, this wasn't much fun to work with, and I didn't like the aesthetic experience very much. For NanoArena, I took the heightmap that was kindly made for me, turned it into a 3D mesh, and then painted that mesh like I was painting a skin on a model. Time-consuming? Not really, for a small map, frankly. For a huge map, it'd get unwieldy, and I'd want to make more use of automatic tools like L3DT to get the basic details right, but meh.
With SM3, I felt like I was working with a bunch of very abstract layers, and my lack of ability to see my results in anything like realtime was a huge hurdle for development of the map.
2. It's gotta, gotta, gotta have a feature-placer that can be done either in Spring itself, or in Hugh's thingie, or whatever. That's a major problem and I dunno what else to say.
3. DECAL PLACEMENT WITH ALPHA. Decals that fade out to nothing with distance... would make a huuuuuuge difference in how this map format feels.
Things that would be nice, but I could live without...
A. Being able to specify somehow the splat angle sensitivity, so that we can have extreme height differences that aren't distorted.
B. Better than that... how's about a Feature format that has less suck? I've been begging for one for ages. Now we're getting it back-doored in through LUA GUI stuff, but I've gotta wonder... will any of that be non-turn-off-able, controlled by the artist, and something that is consensually experienced in sync?
C. Some sort of 3D terrain that is fully 3D, yet works within the 2D pathfinder, that isn't a Feature and is designed to run fast with minimal suck (allowing Features that are designed to be full-featured and can potentially drag down performance quite a bit if used poorly). We've already talked about that, JC, so you know what I wanted, and I heard you say you wanted something completely different. I dunno whether K-Man ever got pathfinding to the practical stage for a Quake-map style world or not, but I guess I'm still waiting to see, and pretty dubious.
I still think that a 2D base mesh combined with 3D objects that are seen as 3D by a pathfinder and sim code would be ideal. Doing things the way that they're currently done, where Features aren't included in the initial pathfinding calc... is a huge waste of computational horsepower, and strongly detracts from this format's ultimate utility.
In short <pants from long-winded ranting> this has great potential, but is not yet where it should be. I will quit bugging people about this format until the more major issues are addressed, now that we've all seen the practical issues, which should make Forb happy. But I'm more than game to come back to this, if the major issues- mip-maps not behaving properly, terrain deformation support being taken out to allow the tesselator to work without so much extra crap, etc., are addressed. In my opinion, this map format should be an alternative to SMD, with advantages in terms of size-of-map, performance, and small footprint going heavily to SM3, but beauty going to SMD. Then mappers with different philosophies can have their genteel arguments about it all, and we can enjoy the resulting artistic chaos
or a ^^ and THEN it's light hearted banter.... 
