Earlier this month we blogged considerably about how Google's Web Accelerator was breaking the Web, or showing which Web applications were already broken. That's because we didn't follow the guidelines for doing a POST when the request has the potential to change the state of the resource. Do you think we learned our lesson? Of course not. In fact, I bet most of you haven't even read the Architecture of the Web. Still hacking. Repeating the same mistakes. Rael has decided that feed:// is the new http://. Web? How many ways may we break thee?
http://radar.oreilly.com/archives/2005/05/feed_is_the_new.html
This is best explained in full. The reason Web Accelerator broke 37signals, is because both teams programmed aggressively. 37signals ignored the SHOULD and Google naive thought the SHOULD was a MUST. If you program defensively, pretend the SHOULDs are MUSTs in your code and pretend everybody else ignored the SHOULDs, then nothing should break (famour last words).
At the moment, I don't see how feed:// could cause dataloss. In fact, I doubt it could. But, I could have said the same about non-safe GETs last month.
Randy