Do you believe that there is only one "right way" to do things? Do you ignore evidence that other solutions may also work? Do you dismiss the possibility that there are reasonable trade-offs to be made between the "right way" and some other way? (do you leave angry comments on my blog?)
Do you think I'm kidding? Read these comments about GET vs POST. There's no question that POST is the "right" way and generally safer, but sometimes it's annoying. Even though GET isn't "supposed" to work, it often can be made to. Don't believe me? Google does billions of dollars a year in GET based transactions in the form of CPC ad clicks (which can cost over $50/click). A lot of other things work this way too, including most non-ajax webmail (for changing read/unread state, at the very least).
If you're writing missile guidance software or something then, please, please, please do things the right way -- it's better safe than sorry, right?
However, if you're writing little web apps, then there's a good chance that you can bend a lot of the rules and produce a better product sooner. I'd rather have something imperfect but useful and popular than something "perfect" but unfinished and unused. Sometimes, it's better sorry than safe.
When is it safe to bend the rules? Just ask yourself, "What's the worst that could happen?" If the worst case isn't too bad, then you're probably ok. Minor bad things will happen no matter what, so it's often better to put your energy into general problem recovery, rather than imagining that you can simply avoid all problems. (this is especially true on the web)