[eluser]jedd[/eluser]
It's all pretty stateless, so yes, anything is possible. I wouldn't imagine this would skew your stats hugely, but if you want to be as accurate as possible I guess you have to cater for it.
You'll have to retain cookies, then. And cope with users that disable cookies on principle (people like me, for instance).
You'll have to come up with an algorithm of what makes you think one user is different from another, as far as your web server can tell.
Your logic for your one-day-apart assessment will be interesting - if you have a cookie, then it's easy - you can store a hash in there and check against it in your logging table. Absent a cookie it becomes more of a challenge - you'd have to do a time delta, if you've established the same IP & browser pairing already existed for that particular page.
So far I see you need something like this:
Code:
logging (
id SERIAL
page -- not sure whether you want to do this by number lookup, or by view name
ip_address CHAR (15)
browser CHAR(100)
cookiehash -- take your pick, probably md5 or uuid
when DATETIME
)