Web Audio API on Firefox vs. liv3c0der
A while back, I've been tweeting about my enthusiasm regarding Web Audio API on Firefox/Gecko and promised a quick fix to liv3c0der which I didn't really deliver up to now. I just quickly wanted to share my latest findings and what this means for using the Web Audio API on Firefox. (tl;dr: Probably you don't want to at this point in time)
So, when I initially did the needed work to adapt liv3c0der to Firefox, I stumbled over some stupid bugs in the LC codebase (or, put more simply: stupid errors I did), but after fixing those, it still had some problems. One of them was kind of unpredictable CPU usage and I haven't yet fully figured those out.
One problem I initially blamed on buffer underrun (and thus the CPU issue) actually turned out to be a more complicated problem in that obviously the DelayNode has some specific problems when used in a feedback configuration. (you can find a testcase here which runs on current chrome, probably does not run on Safari (didn't test, though) and fails quite noisily on Firefox (both 26 and a current Nighly). I've chatted with Paul Adenot about it on twitter, who implemented a lot of the Firefox stuff and he pointed me to #932400, so there is some chance that it might get fixed not too far into the future.
And while it's kind of exciting to being able to help to flesh out the initial problems with this API by using it in this early stage, I am also kinda bummed that this completely ruins my plans of releasing the cross compatible liv3c0der release.
On a more positive note, I might be very well able to make a rather big announcement soon-ish regarding the original cause of this website.