Relying Party Stats as of July 1st, 2008

Month after month we have been revising the filtering used to determine the unique our RP count. We generally uncover several patterns a month, that left uncaught, would cause a single site to be counted as more than one RP. The poster child for such a pattern is blogger.com, as described we described in May. Additional filtering, of course, gets applied all the way back to the first day we launched myopenid.com, causing the entire graph to get revised slighted downward.

However, this month is the exception. Instead of a retroactive downward revision, we have a fairly significant upward revision. How is this possible? This is a very good question, and it had me scratching my head for a bit this morning. It turns out that we were overly aggressive in filtering trust roots that had explicit ports 80 and 443. That is, realms such as http://pibb.com:80 or https://*.jyte.com:443 were not being counted.

Although the number of sites accepting OpenID is still a small fraction of the web, the trend here is obvious.

~ Larry Drebes