aol proxy servers

aol proxy server ip number rotation log analysis incorrect too large damn!

[michaelmuller]

I'm investigating this issue in depth now as my company just upgraded to wt log analyzer 8.0 upon being told the new version gets around aol's proxy server rotating ip numbers without cookies. We now find out that's incorrect.

We have insiders friendly to us (bod) who told us they'd pass on any ideas we have to their development managers. So I'm writing a document.

I believe, with some code changes on their part, that it would be possible to combine the user agent information with the proxy domain name to filter down closer to individual users, and then actually scrutinize the referer pathing in case two users from one proxy server happened to be trolling the same site at the same exact time with the same exact user agent string (which is possible, I guess).

Doable? I've looked at the raw logs and see that

-Michael

[michaelmuller]

Oops... clicked post before I was done.

I've looked at the logs and see that the user agent strings stick with the ip numbers mostly. I haven't started to sig into the proxy numbers, but I have a feeling that unless you get to millions of hits a day on a geographic-sensitive site (such a popular nyc site, where thousands of aol users would be coming in through a small number of proxy servers) it should bring the numbers down closer to reality.

[FineWare]

Depends on what they use for a browser. Some browsers change UAs when grabbing the favorites.ico, and also if there's Java applets (not JavaScript) on any of your pages. So you may have to also correlate the fact that it's an AOL server as well as the time of the hit - but even that's no guarantee if there's more than one person hitting from AOL at or near the same time.