Comments on: Site measurement is hard http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/ The Official Blog of Cogmap, the Org Chart Wiki Sat, 22 Nov 2008 03:45:47 +0000 http://wordpress.org/?v=2.6.3 By: Rob Wilkerson http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/#comment-247 Rob Wilkerson Mon, 29 Oct 2007 19:41:04 +0000 http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/#comment-247 Not any more. Actually, not ever (effectively), but it was a lesson that got learned over and over until it set in for good. Intuitively (and theoretically), IP is such an obvious choice for uniquely identifying users. It would still be useful to do so, I think, but most of us have been beaten with that stick too many times. We know not to bother. We also understand the web and its infrastructure much better now than we did in its infancy. At the time, most of us knew HTML, JS, maybe a little CGI, but the big picture was still hazier than it should have been. Not any more. Actually, not ever (effectively), but it was a lesson that got learned over and over until it set in for good. Intuitively (and theoretically), IP is such an obvious choice for uniquely identifying users. It would still be useful to do so, I think, but most of us have been beaten with that stick too many times. We know not to bother.

We also understand the web and its infrastructure much better now than we did in its infancy. At the time, most of us knew HTML, JS, maybe a little CGI, but the big picture was still hazier than it should have been.

]]> By: brent http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/#comment-246 brent Mon, 29 Oct 2007 13:50:57 +0000 http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/#comment-246 Does anyone ever do that? There are so many problems with trying to use IP as a unique ID that I never even associated the two issues. Tons of companies use a firewall that makes all the traffic look like it is coming from one IP, but to switch IPs from request to request as it passes through the firewall is "special". I assume this is something AOL built as part of dealing with certain scalability and caching issues. Everyone knows AOL caches tons of stuff. Actually, that is something that hasn't really been discussed: Third party caches like these can be real deal-killers. Does anyone ever do that? There are so many problems with trying to use IP as a unique ID that I never even associated the two issues. Tons of companies use a firewall that makes all the traffic look like it is coming from one IP, but to switch IPs from request to request as it passes through the firewall is “special”. I assume this is something AOL built as part of dealing with certain scalability and caching issues. Everyone knows AOL caches tons of stuff.

Actually, that is something that hasn’t really been discussed: Third party caches like these can be real deal-killers.

]]> By: Rob Wilkerson http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/#comment-243 Rob Wilkerson Fri, 26 Oct 2007 11:35:51 +0000 http://www.cogmap.com/blog/2007/10/25/site-measurement-is-hard/#comment-243 The "AOL complexity" with respect to IPs (developers that have been around for a while may have a few other, less euphemistic pet names for it) is well known to most web developers that have been around for a while. For us, though, it's less about metrics and more about security. Want to attach an IP address to a session as a supplemental unique key? Can't do it. Want to filter on IP to allow certain functionality to a particular user base? Can't do it. Want to check that a session hasn't been hijacked using the visitor's IP address? Can't do it. I could go on, but you probably get the idea. :-) Just a different limitation caused by the same problem - the inability to rely on IP address as a unique identifier of a user within a session. The “AOL complexity” with respect to IPs (developers that have been around for a while may have a few other, less euphemistic pet names for it) is well known to most web developers that have been around for a while. For us, though, it’s less about metrics and more about security.

Want to attach an IP address to a session as a supplemental unique key? Can’t do it. Want to filter on IP to allow certain functionality to a particular user base? Can’t do it. Want to check that a session hasn’t been hijacked using the visitor’s IP address? Can’t do it. I could go on, but you probably get the idea. :-)

Just a different limitation caused by the same problem - the inability to rely on IP address as a unique identifier of a user within a session.

]]>