References I found enlightening about the actual practice of it (in favor of it) are the book Ubersleep:
http://www.lulu.com/content/2649551
and Hacker News threads like this one:
http://news.ycombinator.com/item?id=673726
I've read enough anecdotal evidence to be convinced that when it polyphasic sleep works, it works -- that after a potentially rough (and potentially impossible, depending on the person) period of adaptation, the practitioner is not significantly impaired in day-to-day functioning. Skeptics' arguments usually boil down to "it's unnatural", "you can't get something for nothing", or "it didn't work for me".
In terms of needing to set an alarm, I don't know if the body can learn a polyphasic schedule or not; even if it can, the reasons busy monophasic people use alarms to regulate their sleep would apply all the more so to polyphasic sleepers, and I get the sense polyphasic sleep is a sort of "unstable equilibrium"; and if the body can't learn it, that still doesn't mean it's unnatural.
Science could come along and discover a long-term deleterious effect of getting so little sleep, in which case I'd be all ears, but it seems doubtful. I hope the science catches up, though, and we figure out what the deal is with sleep, so we don't have to argue quite so irrationally. There's a lot of interesting recent work on whether being unconscious for eight hours is mostly a behavioral adaptation rather than a biological necessity.
My general philosophy is "if it works, do it", and it definitely works in some instances. Reducing total time spent sleeping per day isn't like smoking cigarettes, it isn't going to slowly poison you. Lots of people get very poor sleep and are definitely impaired during the day, more so than a skilled polyphasic sleeper would report, and they aren't doing long-term damage.