For me RAID 0 was a big disaster. They crashed twice and at that point I got rid of RAID 0. The same drives continue to function flawlessly. Also RAID 0 uses the CPU to do part of the work so you probably loose more than you gain anyway.

CPUs will have to be a lot faster for RAID-0 to be attractive to DV for advanced preview and rendering. Normal use won't fail you with the even most of the 5400rpm ATA66 drives. Nice to have the headroom.

Applications installed on RAID are remarkable, whether the dodgy-for-video card/BIOS based RAID, or the more luxurious WinNT/2k/XP NTFS dynamic disc volume striping. Not that applications take long to load, but I do judge a machine's snappyness by how quickly some of the big apps load. Also, having just applications on a RAID-0, -5 or -10 array doesn't really effect your ability to recover. Making the system drive or "program files" RAID-0 with "NT striping" isn't all that easy though.

If you deal with DV and uncompressed, or multiple DV streams, then more separate drives can be good. Sometimes folk stripe them up, but this only aids housekeeping/selecting the media.

RAID-0 is somewhat out of favour, where a year or two ago it was favourable. It'll probably return, with CPU advances, and possibly when we have shiny discs with so much capacity and high data rate that compressing the media would be berserk.

All of you guys forget to mention what kind of RAID you're talking about: IDE (EIDE) or SCSI. I will never setup IDEs for any kind of array, simply no reason: IDE drives are so cheap and so big, that risk (in RAID-0) and configuration effort (for all other types of RAID) is not worth it, plus, as you all mentioned it will simply use more CPU cycles. Plus rendering on IDE drives used separetely will probably have better performance than in any kind of RAID. So, why bother configuring all that stuff, then praying that windows will not fail drive, somtimes erroneously. Even for very huge project you can have one 200 Gb drive for capture and another for rendering, etc.
SCSI is totaly differen animal, they are sure fast, sure they are reliable, but they are small (compared to IDE) and expensive. Setting up RAID on them will make a lot more sense: for typical project you will end up with at least 2 or 3 drives only for capture material if you wont RAID them. It will get hard managing them when you have couple of projects going. So, my number one objective on SCSI drives is to simply to put them together, to have one single storage area, and not performance. You wont really gain that much from SCSI drives either, though SCSI doesn use any of the CPU cycles (they have their own), but still, performance gain is not that great.
BTW, Serial drives are to be treated as IDE, they are the same drives, just differnt transport, speedwise, pricewise, spacewise, performancewise they are next to IDE. Maybe in a couple of years they will get to the speeds serial is cabable off, but not yet.
All this said, I myeself, still running RAID-5 on one system, RAID-0 on another and simple bunch of IDE disk (BOD) on my third system.

P.S. Software RAID is a total joke. Why would one us it? And tell me, how system with software RAID can be faster than the one wouth it??? With software raid system needs to: calculate RAID parameters, find out location of data, then get it, then process it, when not RAID system simply request data from drive and process it. How can it be faster? It has to do all OS things, plus you want it to do RAID stuff??? The only RAID that ever should be considered is hardware raid, with dedicated RAID controller, of couse if you justify it

I'm running RAID 0 on a pair of WD1000JB's with SATA adapter cards. The thruput is about 50-60 Mb/sec, as opposed to 40 Mb/sec with a single WD drive. There's no real advantage unless you're doing a lot of file copies from disk to disk, like I do for backups. Doing uncompressed avi's, RAID drives are essential, but, I don't do much uncompressed. Still, I've never had a problem with RAID 0. Drives are a lot more robust than they were a few years ago.