The incredible shrinking textures

They are in the thread in detail. I don't want to sound mean, but read the thread.
Hell, I agree LOD is important but we have to take into consideration other peoples systems, the servers and the network. There are certain things we would like but can't have. You are not going to Get Wow level detail here because the dynamic system will crush it out of everyone but the most cutting edge.
Just because you can see it, doesn't mean I can or anyone else.

No worries about offending me, if there's reason to RTFM i definitely will do!

OK, so I reread the thread, but was unable to find actual technical references to anything else than S3 Texture Compression. I really genuinely wish to understand these issues fully, as I'd like to create as high quality as possible without negatively impacting other users.

Is there a spec on the LOD stuff SL does at the moment to be found somewhere?



I think I'm being misunderstood here. My point graceful degradation. I'm not saying that I feel anything over 1024x1024 is even necessary, I don't. But I do feel there is a flaw if the client can't dynamically reduce texture sizes instead of choking.

Educating everyone who is cluelessly creating content isn't viable. Therefore the logical solution is to see to it that people with low end systems won't suffer if large textures are occasionally used.

I think the original Black and White was a good example of an implementation of this. I ran it on a 700MHz computer with a low-end Matrox gfx card. Still, the engine allowed reasonable detail up close, while zooming out a great deal was possible as well.

It is my understanding that mipmapping does in fact increase the memory requirements for a texture. [1] So what needs to be done is to send the video card a smaller texture in the first place. Which is apparently done to some degree at the moment.

I guess I'm just curious about how the algorithm works, and I'd like to know if it adapts when there are more textures shown at the same time. If you're in a crowded room with tens of people it makes sense to heavily reduce detail, then again if you're only seeing one person the opposite is true.

I don't wish to start a flame war here, but I'd really like to see the specs of the algorithms currently employed if they are available somewhere. Primarily to educate myself on getting the best of both worlds, I guess.


[1] http://en.wikipedia.org/wiki/Mipmap