The most frequently retrieved document from this site is my syndication feed. I can tell that from the otherwise lousy statistics provided by my ISP. I also use Google Analytics, and the crude statistics from FeedBurner. Analytics doesn't tell me anything about the use of the feed, since I cannot embed javascript in a feed. Only Googlebot (I use it as a Google Sitemap) and Feedburner accesses the feed directly. The rest of you get the feed via Feedburner.
Feedburner keeps a cache, and refreshes that every 30 minutes. The feed is retrieved around 250 times a week, and I would have expected 2*24*7=336 hits on the feed. Note that 250 includes Googlebot. Since I've no direct access to the Apache log files I cannot tell how the file is retrieved. If I was Google I would add an If-Modified-Since header to my GET request and looked for a Status 304 (not modified) in the response. Then I could get away with a much lower access frequency and gain throughput. This would explain the slightly lower access rate than expected.
My estimate is that about five to ten persons visits my site daily and that two or three of them read something. Furthermore, between ten and fifteen readers follow me using the feed. In particular the use stats for the feed is hard to evaluate.
All the entries on this site are written directly in Atom using nxml mode for Emacs. The site is generated from this source using a number of different scripts, the bulk of it written in XSLT. The source file for this entry is here. There are two things you have to think about when you write and design your content for syndication:
The problem with javascript manifests itself differently.grazr.com has a nice category browser, which can be used on some of KB's data. It seases to work at least when stuff is syndicated on Google Hompepage.
Fig. 1. My previous entry as it appears on my iGoogle Homepage. Note that all the captions appear next to the first two images, not as they should adjacent to each corresponding one. This is Google Chrome but it looks the same in Firefox.
When I write these entries I preview them using the server I run on my laptop. Making the links absolute is the last thing I do before publishing to the web. I forget this now and then, and I won't see that before it has been through feedburner and reached google and bloglines. The rebuild-run-debug cycle is long for syndicated content.
There might be problems with the feed readers' HTML support. I'm currently only the online ones, iGoogle and Bloglines, so the html shouldn't theoretically not be a problem on them, but still there is.
One is that (for instance) Google defines stuff in its CSS which changes the behaviour of tags in a way such that my content does not look as intended. The <kbd>...<kbd> (keyboard) didn't work the last time I checked. The other problem is that (for instance) Google actually edits my markup. It seems that in my previous entry, the one about stanford, Google stripped the <br clear="all"/> elements from its attribute (clear="all") and that CSS redefines the break tag in some way or another (see Fig. 1).
Please bear with me. Below you find the content from my previous entry. In this version I'm employing some other methods for achieving a break. I indicate in the text which method I use.
A lovely garden somewhere between our hotel and the Meyer Library at Stanford University. Break is the original <br clear="all"/>. Predicted behaviour: No break will be generated in Google.
Gustave Rodin at Stanford University. You may want to compare this with the Rodin in Copenhagen. Break is using CSS <br style="clear:both;"/> Predicted behaviour: Will not generate a break if Google strips all attributes on <br/> and generate one if it just strips clear
A nice graffiti, or is it al fresco? And what is the difference? Break is using an empty <div style="clear:both;">...</div>. Predicted behaviour: Will generate a break in Google.
Science against an evilgelical religion. Break is again a <br style="clear:both;"/>
The Papua New Guinea Sculpture Park. <br clear="all"/> as before
The view from my hotel room.