okkkkey.... I made many tests, and now I have a big headache, so I'll stop it for the mo.
here are the results:
I solved my problems with the neb bitmaps and I was able to try both pcx8bit and jpg.
I noticed that the background nebs are rendered over a flat plane, a square.
When fs2 wasn't able to find the images (don't know why, after a reboot everything went normal:confused: ) it rendered this plane as full white:)
The square dimensions are those setted in fred.
BTW this is pretty crap in my opinion, since you can't have big images surrounding you: being the square flat, if it is too big it generate a weird fisheye effect at the borders. You don't notice with the simplistic vanilla nebs, but when the image is more complex it is weird.
On the other hand, you don't have the warpings caused by the sphere model, even when the sphere is high poly.
BTW probably those warping on skyboxes may be solved using something different than planar uvmaps, but I'm just tooo lazy to search a solution
Also I had big performance issues. When the 8bit pcx was fine, the jpg version (almost same order of kb size btw) killed the framerates. I remember RT saying that there was a bug with the system rendering twice the TGAs. Maybe it is the same problem? I'm using 1_20 version.
The highpoly (IIRC more than 500 polys) skybox with the same image + a second 1024 image was causing about half the performance hit of the background bitmap version. Since I used jpgs too, the problem could be not related to the fileformat, thought.
Also I had another bug with 1_20, which never noticed in older versions: when I move the skybox "shakes", it doesn't stay still (the cause of my headahce:p).
Stars:
Personally I think that a star image like a nebula must have an higher concentration of stars on it compared to the deep space areas. Also, I may want for example to create a milky way, or a spiral galaxy, etc.
I tryed to remove the star layers from the images, but didn't like the result.
If I put my nebs with stars in game as background nebs, the stars aren't that good obviously. They aren't crap at all, but viewing stars without motion blur only in a specific area isn't all that good.
In a skybox instead, reducing at minimum the number of stars generated by fs2 and using a coherent starfield image tiled over all the model except the nebula area works fine in my opinion.
There still is a difference between generated stars with motion blur and stars in the textures, but I noticed it just because I was aware of it since the stars in the textures are diffused all over the skybox. Also it gives a depth effect, with the stars with motion blur looking like if they are closer than those on the textures, which could be nice or not, personally I don't dislike it at all.
A fair solutions btw would be to have as minimum star density in FRED 0% instead of 10%