I've been prototyping a few pages that use webkitspeechrecognition. I learned quickly that you cannot load these from a file, you have to serve them from a webserver. I'm using osx so I just moved my files to the local apache that was already running and enabled. This worked fine for quite a while.
For some reason, none of my pages that were working fine will prompt me to deny/allow the microphone usage. I even copied an existing working page from another webserver and if I load it from http://localhost/speech.html it will not prompt. It skips the prompt and goes to my recognition.onerror handler and logs "not-allowed"
However, if I load the same page (or any of my other prototypes) from http://127.0.0.1/speech.html it works fine. This made me think I had accidentally cached a response like "always deny" or something. I think I cleared/reset all my chrome settings but I'm still getting the same behavior. 127.0.0.1 will properly prompt, but localhost will not prompt at all.
Where might chrome be storing some additional settings that I need to clear?