The app I'm working on for a living was brought down recently because FB changed the name of a method in their javascript library.
Once the origin of the bug was discovered it was easy to fix, but that's still a pretty unreasonable thing to do when, as was the case, a major library that is linked to from the very front of the wiki (the FB/AS3 bridge) is broken by an update.
It's also pretty painful when the median call to a certain specific API method is about a second or two, but the peak is literally ten seconds.
Someone was editing the code on the live site and made an undocumented change.
The call I'm talking about is a javascript call made through a Flash app. The problem isn't that a page is timing out, the problem is that we have to wait for this call for the app to allow the user to continue.
Well done if it's working for you though, hope it continues to.
Facebook's API does not have the typical warning labels about shifting API calls, development releases, or technology previews; these are the final, real versions, and they're expecting people to just roll with it when they whimsically decide to rename an API call or database field. If you expect people to just roll with that kind of thing, that's cool, but you have to provide fair warning -- you can't just recruit tens of thousands of corporations to use your software and then go making API-breaking changes willy-nilly.
The open-source community, where _all_ development at every stage usually occurs out in the open, takes stability, etc., very seriously. It's good and important that early versions of software are available, but they're tagged as such; ext4 had been in a stabilized "experimental" mode for the last two years or so, and it was relatively stable during that. Very early work on Firefox is available, but it's not pushed out until it has a year or so to mature, and it's billed as alpha and beta, despite the fact that the Fx 3 betas were rock solid, as are the 3.1s. When KDE released 4.0, they gave everyone fair warning: despite the name, it was still an experimental platform and release.
This is serious business. There's nothing wrong with releasing early and releasing often and sharing your contributions and collaborating -- that's good, and would probably help Facebook a lot if they would take it to heart. The bad thing here is that Facebook is actively promoting and encouraging new users to adopt this platform for incorporation into live, production-level applications, and then breaking the API and engaging in other assorted mischief without notice. That's horrible. :(
The common mantra is to release early, release often. Now I admittedly don't know how often FB is releasing, it seems like the crux of the complaint that they released too early... and the fix isn't to release more often (at least at the author's implication.)
Would he honestly rather they hadn't released anything at all? How would you feel if this were your app? I mean, the critiques seem well-formed, I just don't know if it's symptomatic of having waited for fixes that aren't coming or not.