Colin,
I agree with you 100% if you are referring to software raid, but h/w raid should be totally reliable and has obviously demonstrated its stability within buisness environments as very few people would opt for a non raid soultion.
I am lucky enough to have picked up an ATA Raid card with 8 ports from a server which was being decommisioned and have used this with no failures since the very early days of pluto.
NOS
It is in fact hardware RAID's in enterprise environments I am talking about! People set far too much stock in these. Until recently I was the Regional Operations Manager of a multinational for the Asia Pacific region, covering around 6000 seats, running over 350 servers in everything from DELL PERC RAIDs, IBM ones and EMC DMX/Clarrion SANs, DAS, NAS, iSCSI and Centera's. I was ultimately responsible for all failures in the region. Believe me, they fail (broken RAID sets, split hypervolumes, failed RAID backplane hardware, spontaneous degrading, and inconsistencies between the RAID container metadata and RAID BIOS metadata during reboots that cause a RAID config to be completely lost - if you don't catch this one before any new data is written to the drive, it is almost unrecoverable!) far more often than people think, or than they should!
I'm certainly not advocating businesses not use RAID (HW), I'm just saying with the issues above combined with endemic oversight of RAID health monitoring with IT engineers, often RAIDs merely increase availability/MTBF rather than actually provide real "redundancy"! (its amazing how often an engineer came to me with a credulous look on his face and explained that 2 disks failed at exactly the same time - to which I always replied "bullshit!" show me your RAID health monitoring for that set for the last 3 months - of course they couldn't!) Further, I'm certainly not making a case for HW vs SW or the other way around for home systems, but I will say that some of the only systems I have worked with that actually provide real "redundancy" were software systems such as those employed on the EMC Centera, which are truly able to "self heal" rather than dumb availability from RAID - but those are insanely expensive
(very clever tho!)
btw, in a business environment you should never use RAID5 anyway, it is most definitely the "ugly sister" of the RAID family, very slow and unreliable for the sake of saving a few bucks. RAID 10 is usually the way to go, and if you can hyper/meta-volume them as well, all the better.