And this really does need to go on the front page because far too many startups are using Facebook Login, and the recent partnership with YC to get YC startups hooked on the FB platform.
What a fantastic explosion of buzzwords and slightly bizarre assertions.
I know you interned at Facebook and are familiar with the site architecture, but the majority of the things you've mentioned here had absolutely nothing to do with the last outage, and it's far too early to draw any conclusions about the current unavailability.
This sounds interesting but you're using a lot of jargon, could expand on it?
I know the thundering herd problem is when a resource goes away and then everybody tries to fetch it the moment it comes back up, sending it down again. What is the subscription update model you're referring to, and to what portion of the system does it apply? I was under the impression that user stream updates were distributed somewhat asynchronously, and were heavily cached, but I could be wrong.
And.. why is it down now and not when I was working? At least, it helped me become a bit more productive last time. This time, it's just ruining my night. ;)
I wasn't really asking if it is down. When I submitted the post, I already confirmed that it is down. This is a major problem for the whole internet now that a lot of of websites use Facebook Connect and other APIs. Some businesses run on Facebook. It's a news. Definitely tech related. Definitely interesting for HN aficionados.
To be fair, people stopped blaming Rails for Twitter a while back, after it became clear that the problem was mostly at the data level.
Likewise, nobody (well, except liuliu) is blaming PHP for Facebook's downtime, since last time they blamed it on a poorly-handled failure case after an accidental misconfiguration:
Facebook rolls our code changes independent of product releases. They just need to flip a configuration switch (usually for a small percentage of users at first) to launch a product, not release code. This can still cause site issues, but usually just some error message, not the entire site being down.
This does not look intentional. If they were just rolling out a front-end change, they wouldn't have killed their APIs like Connect and the Like buttons.
Really no need for this to go on the front page though...