$_SERVER['REMOTE_HOST'] and proxies.... help!

nbagley

I am having a interesting problem... if any one can help me (or just throw any idea at me) it would be of great help!

I am tracking remote users ($SERVER['REMOTE_HOST'] or $SERVER['REMOTE_ADDR']). Not a problem. The problem that I am running into is proxy servers. I am getting a single user registaring from multipule IP address, like this:

cache7-5.ruh.isu.net.sa
cache1-2.ruh.isu.net.sa
cache9-5.ruh.isu.net.sa
cache10-5.ruh.isu.net.sa
cache2-2.ruh.isu.net.sa

I wouldn't normally have a problem with this, but I need to be acurate in my tracking... these are all posting to my database as unique hits to the web site when they are really session hits (or hits orginating from the unique hit).

The problem looks to be that the web sites files are cached on the proxy server and thus pings my database from the cashed location instead of the server that the website is located on.

The application that I am building is a thrid party add on to a site. I can not force a web site to not allow cashed pages to be stored. Can anyone think of a way around this bug?

Like I mentioned... anything would be of help to me. Thank you!

Mordecai

Hmm... ignore the first part? Of course, that could limit some people down to just their ISP, which wouldn't be accurate...

You could try using the Levenshtein distance algorithm to try and cancel out those that are very close (within 4 letters of each other or so).

if (levenshtein($remote1,$remote2) <= 4) {
  // $remote2 is very close to $remote1, don't add to db
}
else {
  // add to db
}

Of course, how you get a list of close addresses is up to you... I'm sure you can just use the last 10 or so letters (or base it by ISP or domain).

nbagley

Thanks for the thought...a good idea, but I don't think that will really work the best.

Think about this: There are two seperate users come to the web site at the same time, and visit the same page (which happens to be located at cache7-5.ruh.isu.net.sa on one of the proxy servers). By studing the similarity between the two hits the Levenshtein algorithm still won't know that two unique hits just occured. Doing it that way, the program will assume that it's the same user (i.e. maybe just refreshing the page).

Basically what that would do would be limiting one user to come from that specific server in the amount of time that I have set for a unique hit (i.e. every 15 minutes). Which, as you probably have already denoted, stinks. Accuracy goes out the window and visits the problem that I have right now.

Welcome to my nightmare. Thank you so much, though, for taking the time to think about it. It seems that I always get the weirdest questions that no one knows the anwser to. If you happen to have a stroke of enlightenment again, please let me know. I would be endebted. Thanks again.