Idle Time

Simone Manganelli, on his blog Technological Supernova, which I apparently link to a lot (just kidding!), presents a breakdown of pages that link to his blog. So, always one to jump on a fun-looking bandwagon, here’s my implementation of the same thing…

First, all stats I’ll present here are from my blog for March 2007. They don’t count the rest of my site, they don’t count the old blog, and they don’t count any other time periods.

Also, I don’t use Google Analytics; I use Webalizer. Webalizer only tells me most numbers in terms of requests. A request isn’t necessarily completed; the connection may be aborted, or the response may bear a status code other than 200 OK. There were 139,446 requests in the month of March, of which 83,811 were completed. This is roughly 60.1%. Keep this in mind. For the rest of the article, I’ll use the real numbers of requests; I won’t calculate the completed-requests estimates.

RSS feeds account for a crushing amount of traffic:

• 1. /blog/feed/atom/: 32117 requests
• 2. /blog/feed: 13214 requests
• 6. /blog/feed/rdf/: 5375 requests

I already intend to move the feeds over to FeedBurner at some point, but I haven’t gotten around to it yet. I’ll definitely see my bandwidth usage go down when I do. The feeds totaled 50,706 requests for the month.

The reason why feeds get so much traffic is because most RSS readers (I extrapolate from Safari and Vienna) default to polling the feed every half-hour. If we assume an 8-hour workday during which you’re checking my feed, that works out to 16 requests per day, or 480 requests a month (496 for March). That’s per-person. Allowing some variance for some people having their RSS reader open for more than eight hours, I think we can assume that I have somewhere between 75 and 100 subscribed readers. (Hello, all of you!)

Various housekeeping files account for a lot of traffic, too. commentPreview.js is #3 with 9941 requests; quoter.php is #4 with 9910 requests; the blog style-sheet is #5 with 7708 requests; the comment-posting file is #9 with 751 requests; and the XML-RPC file (for trackbacks) is #17 with a paltry 293 requests. These files totaled 27852 requests—almost as much as the Atom feed, and roughly 9000 more than the other two feeds together.

So, what are people actually reading? Here are the top ten.

Most popular URLs

• 7. LMX 1.0 released: 4081 requests
• 8. The main page: 2207 requests
• 10. Screencast frame-rate tips: 691 requests
• 11. The first working version of Negative Turing Test: 576 requests
• 12. LMX and Adium message history Q&A: 512 requests
• 13. Compact archives view: 473 requests
• 14. The #pragma mark movie (yes, the movie itself): 418 requests
• 15. Mac app checklist: 356 requests
• 18. Adding chat styles to Skype: 276 requests (too bad that doesn’t work anymore)
• 19. The announcement of image’s release: 269 requests

And who’s referring to all these URLs?

Referers

The top three referers are direct request (bookmarks, IRC/IM/email links, etc.), with 88,049; other pages on my site (e.g. going from the front page to the article page), with 43,741; and the Goog, with 2638. After that, we have:

• 4, 18, 19, 20: reddit (LMX 1.0 released), with a total of 900 hits exactly. In order, the referrers are the front page of the programming subreddit, with 783; the article page, with 40 JavaScript-less people; the reddit hit-counter redirector (which a story link changes to when you click on it, if you have JS enabled), with 39; and the “new” page of the programming subreddit, with 38.
• 5, 14: Daring Fireball Linked List (LMX 1.0 released), with a total of 754. The front page of DF was most of these, with 693; the front page of the Linked List specifically was 61. Other DF pages didn’t crack the top 30.
• 6, 13, 28: Cocoa Blogs (the front page (249); LMX 1.0 released (93); screencast frame-rate tips (260)), with a total of 600 requests exactly. (The discrepancy in numbers comes from the fact that I only listed here the pages that it mentioned in March, while Webalizer counts everything, including hits to some of my older posts.) The front page is #6 with 482; the front page with “www.” prefix is #13 with 97; and the “all posts” list is #28 with 21.
• 7: Google Reader, with 304.
• 8, 15: Google Image Search. Most people find my frame-grab of the Hillary Clinton Scowl, from my fisking of the SOTU 2006. The US Google Images brought 297 requests (whatever image they found); the UK Google Images brought 60.
• 9: Tim Bray’s link to the LMX 1.0 announcement: 174 requests.
• 10, 16, 17: The Adium blog (LMX 1.0 released; LMX and message history Q&A, with a total of 250 requests exactly. The referrers are the front page of the Adium blog with “www.”, with 146; my link to the LMX-and-message-history-Q&A post, with 56, and the front page of the Adium blog without “www.”, with 48.
• 11: Bloglines, with 122. Seems like Google Reader took their market away from them—Google Reader was born in 2005, whereas Bloglines was born in 2003, but look at the market share (at least on my little corner of the web): GR has almost two-and-a-half-times as many requests to my site as Bloglines does. (Yes, I admit that I haven’t researched this beyond the request counts. It’s just an aside.)
• 12: Fucks per programming language and license, with 115. This one’s an anomaly—a bug in MSRBOT that caused it to misreport referers. It’s a somewhat interesting article, especially if you dig statistics, but it doesn’t link to anything on my site—in fact, MSRBOT claimed that page as the referer on everything it tried to slurp from my site.
• 21, 24: Mark Pilgrim (LMX 1.0 released), with a total of 70 requests. 36 of those came from his front page; the other 34 came from the linking article
• 25, 26: TUAW (screencast frame-rate tips), with a total of 58 requests. 32 came from the article; the other 26 came from the front page.

Finally, a quick look at search strings:

• 64 people found me by searching for Hillary Clinton (see above entry for Google Images among the referers). Of those, 15 managed to do it despite searching for Hilary Clinton (with one ‘l’).
• 14 people found me by searching for “bored”. My pseudonym lives on.
• 14 people found me by searching for “calloc vs malloc”.
• 11 like-minded people searched for “how to eject the dvd rom on mac using terminal”. What are the odds?
• Finally, 10 people found me in their quest for better “quicktime performance”. Presumably, they found out how to enhance it.

Simone mentions that he got linked to from a secret Apple server, evangelism.apple.com. I’ve had a couple such private links in April: two from Adobe’s “kajigger” system (which I guess to be some sort of internal del.icio.us-esque system), and one from Blue Lava Technology’s private wiki.

User-agents

My demographics seem to differ greatly from Simone’s:

1. 22,245 requests (15.95%) are from NetNewsWire.
2. 22,203 requests (15.92%) are from WebKit browsers such as Safari (I haven’t yet worked on making it figure this out more specifically).
3. 18,540 requests (13.30%) are from Netscape (not including the open-source browsers).
4. 12,978 requests (9.31%) are from Internet Explorer. I must have a lot of fans in the corporate world. None of them seem to be using IE 7 to subscribe to my feeds, in case you’re wondering.
5. 12,552 requests (9%) are from NewsFire.
6. 8353 requests (5.99%) are from the MSN crawler. Bad MSN!
7. 7877 requests (5.65%) are from the Syndication framework (i.e. Safari RSS). I don’t know how that separates from the main Safari stats.
8. 5180 requests (3.71%) are from the Yahoo! crawler. Bad Yahoo!
9. 4587 requests (3.29%) are from libghttp 1.0. This is the GNOME HTTP-client library.
10. I’m sure you were wondering, Jeff: 3525 requests (2.53%) came from Vienna.
11. 3209 (2.30%) came from Gecko browsers (Firefox, Camino, etc.).
12. 2684 (1.92%) came from Opera.
13. 1720 (1.23%) came from Googlebot.
14. 1514 (1.09%) came from Google’s Feedfetcher, which is apparently the write end of the Google Reader pipe.
15. 828 (0.59%) came from NewsGator (presumably the online version).

Categories: This blog. | Comments: 3 (feed).