Looks like they were smart and are using Rackspace for the main github and AWS for the status site. That's 101 for status sites.<p>I slightly wonder if it's worth putting status sites in a separate domain as well, for resilience against DNS outages or misconfiguration. Obviously you'd also want status.example.com for example.com too, but having some standard way for services to report uptime (to users and to machines, including deep API information, not just "can I ping it" would be nice.<p>It would be nice if someone set up "arbitraryservicedowntime.net" and did things like github.com.arbitraryservicedowntime.net". But then you'd need some way to associate and promote the independent URLs. Just doing github.com and github.net might work, if github.net is otherwise unused.<p>Pingdom kind of solves this for just ping, but doesn't work for deep status, causes, ETTR, etc.
I'm a bit clueless about network resilience, but doesn't using the url status.github.com imply the status and main website share a DNS service? We have seen name resolution problems in the past so would a DNS problem that affected github.com would also affect status.github.com?
They link to <a href="https://status.github.com" rel="nofollow">https://status.github.com</a> which doesn't work (you have to go to <a href="http://status.github.com" rel="nofollow">http://status.github.com</a>). Bit of egg on their face, sadly.
There's two posts on status pages on the HN homepage right now. This and Balanced:<p>- <a href="http://blog.balancedpayments.com/status-page/" rel="nofollow">http://blog.balancedpayments.com/status-page/</a><p>- <a href="http://news.ycombinator.com/item?id=4877643" rel="nofollow">http://news.ycombinator.com/item?id=4877643</a><p>The hardest problems to me are properly monitoring and computing downtime. Heroku's uptime calculation uses time[1], which breaks down for an API where it should be more about the number of requests.<p>Any idea how Github is measuring downtime? Is there a difference in Github's measurement for when there's low usage vs. going down during peak hours? How are sporadic errors computed vs. a complete outage?<p>[1] <a href="https://devcenter.heroku.com/articles/heroku-status#uptime-calculation" rel="nofollow">https://devcenter.heroku.com/articles/heroku-status#uptime-c...</a>
I miss the whimsical nature of the previous status page (it seemed to fit with GitHub's "voice" better). But I'm guessing this version isn't going to crash when GitHub does (<a href="https://github.com/blog/1261-github-availability-this-week" rel="nofollow">https://github.com/blog/1261-github-availability-this-week</a>). Hopefully it's statically generated.
Do you have negative letter-spacing on large counters? "662ms" under "98th perc. web response time" has virtually no spacing between the glyphs. Looks messed up, border-line illegible. That's on Mobile Safari.<p>Looks great overall though. Nice job, guys.