Harrumph
Technical Difficulties, emerging over the weekend, have taken the wind out of the sails. I want to go back to bed.
First, the service the tracks visits to the site appears to have an idiosyncratic understanding the of the word "Unique," because the list of "Unique Visitors" on any given day shows multiple listings for some addresses, worst of all, my own. It's as though I'm loading the site to inflate my stats! This is the very last thing that I would do (seriously), because I want the numbers to mean something. I could, presumably, write down all the IPs of daily visitors, scratch out the duplicates, and so arrive at a an accurate total, but I can't be the only one to regard this as the tracking service's job, not mine.
Second, there seem to be some problems in the Comments department. A visitor wrote to tell me that, when he clicked to post a comment, he was confronted by an admonitory screen that scolded him for having given offense in the past and that refused to accept further comments from him. This was nonsense; the comment went through. I believe that I know of one prior incident of this, but now I have to look into it. Time to open a ticket at MT. And, while we're talking about comments, I understand that the "Preview" option isn't WYSIWYG.
Comments
That is the extent of your philosphical musings: "What you see is what you get?" I am disappointed.
Posted by: PPOQ | January 31, 2005 11:17 AM
You have discovered the ragged underbelly of website tracking. The fundamental issue is that the protocol upon which the Web is built is "stateless," i.e., the server treats each request for any object -- HTML page, image, document, etc. -- as a distinct network connection. As far as the server is concerned, there is no such thing as a user session, just a bunch of requests from some clients (browsers). Hence the attempt -- which no longer works very well -- to identify unique users by considering distinct users to have their own IP addresses (e.g., 12.0.4.5) or hostnames (dialup24.nyc.panix.com). ISPs now conserve IP addresses by assigning them dynamically -- dialup users are almost certain to have a different IP address each time they connect; cable or DSL users may change their IP address once every few hours, days, or weeks, depending on the ISP; only geeks like me who pay extra for a static IP address will always show up as coming from the same place. (And some geeks will obscure themselves behind various proxies or anonymizers.) Additionally, most corporate users these days are behind a firewall or proxy, so the huge pool of company users will appear to come from only one or a handful of IP addresses. This remains the traditional HTTP log method of user tracking, as ineffectual as it may be.
The problem of tracking distinct users has largely been solved by a reasonably bailing-wire-esque workaround: placing a cookie on the user's browser, containing some kind of unique identifier, which is sent back to the server on each subsequent request. This then allows some logic on the server side, usually with the aid of a tracking database, to assemble the notion of a user session. However, none of the out-of-the box log-tracking software does this without a fair amount of extra configuration; most non-high-end hosting companies aren't likely to offer it as a service to their customers, and if they do, it may tend to be expensive. I once worked on a project that ended up using cheap monthly service that was quite easy to plug in and worked well enough. However, $50 per month may be extremely cheap for a business, but it's non-trivial for a personal website. (With some research, you can likely find a cheaper one.)
I personally keep thinking I'll write some more sophisticated tracking stuff into our site, but other interests (baby, wife, violin) always end up taking priority. So we just generate crude HTTP logs and look at those every now and then.
Posted by: Max | February 1, 2005 11:04 AM
Thanks for the enlightenment, Max; it clarifies the response that I received from Extreme Tracking. I see now that for the numbers to mean much of anything the number of visitors has to be high enough to make repeat visits (not to mention servicing uploads by the author/owner) insignificant.
Posted by: R J Keefe | February 1, 2005 01:22 PM