Did a Mac User Just Subscribe?

Posted by Hans de Ruiter

Looking through the log generated by the web-server, I noticed that a user agent called Apple-PubSub has been accessing blog/rss/, which is the RSS feed for this blog. Apple-PubSub is the name of Mac OS-X's built in RSS reader. If you are the Mac user (or one of the Mac users?) that subscribed, welcome; if you are not, welcome too. This is an example of the wealth of information available in the logs.

Whilst this may sound a little scary to some, there really is nothing to worry about. There is nothing sinister going on; nor is it a case of big-brother watching. In reality, it is similar to a shop-keeper knowing who has visited his/her store. A website owner has just as much right to know who is visiting his site as a shop-owner has the right to know who is visiting his/her store. In some ways a website owner has more information, since the logs tell exactly which pages a visitor has read and when. On the other hand, unlike a shop owner, there is no record of the visitor's face, nor any information that truly identifies who a visitor is. Via various techniques, a web-user's location can be located to the country, or city level - as shown in the ClustrMap, for example. Identifying individual users is unfeasible. 

Given the volume of information available, I think that web-server log analysis has a long way to go still in terms of extracting all the useful information and presenting it in a meaningful way. All of the log analysers are capable of plotting trends. However, sometimes individual case studies could be useful. For example, tracing how a single (anonymous) user navigates through the site could highlight possible methods of making such navigation easier, thus improving the viewing experience of all visitors. All the information is in the log, it just needs to be extracted and presented effectively.

I am currently evaluating three server log analysers: Google Analytics, Webalizer and AWStats. Urchin would be on that list too if it didn't cost $4/month, which would effectively double the cost of hosting. More data is still needed, but each has their own pros and cons. I will discuss these items in more detail once I have used the analysers for at least a few months, but all of them seem to miss one feature that I thought would be elementary: providing the number of downloads of every file. In particular, knowing how many people have downloaded the sample code that I have produced, would be useful.

Comments (2) | 30 June 2008