What is the nature of your survey. I can imagine only a few situations that you would ever need to have one person vote once and only once.

Of all of the methods listed above, the only one that will come close to getting the job done is the one vote per user... of course, a single person could sign up for multiple user accounts and so you have been foiled again.

You could go the route of trying to uniquely identify the computer via a hardware diagnostic (which is a clever idea by the way ... oops - that's a little to close to what a common new operating system does). However, this doesn't prevent somebody from going to a library with 20 computers or a university with 100 computers and voting once from every computer.

You could try a combination of the things. But again, this would make it harder to do, but not impossible. Even if you asked for a unique social security number that you verify against a database, a person could use his childrens numbers, friends, etc.

So the question is, what is the nature of your survey. For most of the time, doing an IP check is "close enough." Yes you may block out a number of people under the same firewall. If it is a broad internet survey, then this really shouldn't matter (show a polite message saying that the class C they are on has already submitted an answer -- yes, use the class C as most dial ups still should put a person on the same class C).

If you really need a single vote per person and critical data depends upon it, ask for a registration fee, social security number and make them sign a contract (via fax) that is prosecutable by law. Or maybe you should do your survey a different way (such as live and in person). If it really isn't that sensitive, be happy with the class C IP.

As we always used to say... "Close enough for atom bombs and hand granades."

my @a=qw(random brilliant braindead); print $a[rand(@a)];