Between midnight and 11:07 AM on Saturday August 2nd 2003, ongoing’s RSS feed was fetched 4,512 times, which is 6.76 times/minute, i.e. about once every 8.8 seconds. Add it up: that’s just over 128 MB/day. This weblog isn’t even in the top 100, if you believe Technorati. Yow. [Update: Follow-up questions from Bill Seitz and Brent Simmons]. [Update Aug 6: Fixed bug, better numbers]
Bill asks “Can you measure how many unique IP#s that was?” and Brent asks:
Tim -- do you mean the RSS feed was requested 4512 times or that it was downloaded 4512 times? If it was downloaded 4512 times, how many times was it requested?
What I'm trying to get a handle on is how much bandwidth is saved by servers that return 304s.
Just now at 4:43 PM there have been 7,138 requests from 844 unique IP addresses, of which 6,937 returned the data (code 200), and 201 returned a not-changed signal (code 304).
Caution advised: everybody who goes at all deep with the Net knows that “unique IP addresses” is a shaky and unreliable way to measure anything, with many sources of both positive and negative error.
Update! · So basically, those numbers don’t make any sense, why aren’t there more nothing-changed 304 signals? It was my fault; the code that changes the picture-of-the-day every ten minutes or so was also as a side-effect re-writing the RSS. So I fixed it, and here are the stats for August 5th, on which ongoing was updated only once:
11,526 GET requests, of which 8,901 returned the no-change (code 304) and 2,625 returned the data (code 200). There were also 359 HEAD requests. These requests came from 1,374 unique IP addresses.
This change is going to save me a noticeable amount of money every month. But, that's still a lot of traffic for a moderately popular blog.
Hm, those numbers suggest that there are a few RSS scrapers that are not doing If-Modified-Since requests. Seems to me that a search-and-destroy mission is in order.