rugby471: Plymouth depends on KMS, but KMS is needed for many other things, like working suspend/resume, BSOD (or any other color hehe) for kernel panics, fast VT switching, to run the Xserver without root privileges.
To answer the other question, KMS is in karmic now but only for intel. Nouveau wont make it, 2.6.32 kernel at the earliest I hear, doesn't support all chips (might be wrong on that point), and nvidia binary drivers are incompatable with KMS. ATI depends on if it gets merged in 2.6.31 which it most likely will. That leaves *alot* of cards that wont have KMS support by the time karmic is released, don't forget about everything that isnt from those 3 companies
Considering that text boot takes roughly the same amount of time (seriously) as the usplash or any of the other modes seem to (The time doesn't come from doing pretty things to the screen or just raw text- it comes from the operations being done to generate either, which have traditionally been done serially throughout the boot up, when there's quite a few things that could have been done in parallel...) your remark isn't actually helpful...
This might not be true in all cases. Got an EEE recently and (for obvious reasons) I decided to replace the sys that came with it. With my own, largely based on LFS-stable but deviating from it when I thought appropriate. Got Plymouth running nicely, but...
As I said I deviated from the official LFS quite a bit, writing custom bootscripts among other things, resulting in an ~8 sec boot time (GRUB to SLIM). Plymouth added 5 secs to that boot time, so I ditched it and went without bootsplash. I'm quite sure that on a recent multicore desktop that delay would be significantly shorter (not crazy enough to risk hosing my main machine, so no experience with that), but it _is_ a delay in this case. (Well, yes, I might have just misconfig'd stuff.)
I don't believe there's any technical reason for a graphical boot to take longer than a text boot. The people designing the graphical boot simply need to know how to optimize what they're doing instead of just relying on faster hardware or an accelerated video card (something that isn't too common in free software outside of the video apps and libraries like mplayer and ffmpeg, and perhaps the most popular X11 drivers).
I wrote some applications for the Linux framebuffer a long time ago (including a boot splash that never got released), and in my testing, on a ~200MHz CPU (IIRC), with a plain PCI video card, I was able to push over 30fps full screen to a 1024x768 screen (that's about 68MB/s with packed 24-bit pixels, which was the bandwidth limit of the video card, not a CPU limit). A boot splash doesn't have to draw to the entire screen, and doesn't have to load a bunch of large graphics files from the drive (if they design it right). I haven't used Plymouth (I mostly use Debian derivatives and Gentoo), but my guess is that its bottleneck is either a lack of I/O optimization (reading too much fragmented data from disk) or not-yet-optimized processor and video resource usage, not an inherent failure of the concept of a graphical boot splash.