I'm not talking about single occurence of the "bad" pattern. If you multiple it to the occurence in a whole code base... well you get the point. That's why underscore and coffee-script treat default assignement that way.
Interestingly, in JSPerf, sometimes the ternary pattern is actually faster than the if statement. Must be something V8 is doing.
you may have to check for null if you want to use a default arg2 while providing an explicit arg3. i guess you can always do:
myfn('a',undefined,'b');
I could see, using the if statement form, that you might have a programmer decide to compact it and drop the brackets - and then you might fall in the trap where somebody thinks they're in an if block when they're really not. Happens pretty often.
In addition, I don't know if you really would want to check for null. I usually just check for undefined, which would be the absence of any provided value (as opposed to a provided value that just happens to be null). Of course, it depends on the app.
In addition, I really don't like using short-circuited boolean ops for assignment. The style is grating to me - it's not immediately obvious what it's really doing. To each his own.
EDIT: Created a simple jsperf test. Difference is about 6.8%, but we're talking hundreds of millions of ops/s. Not the kind of stuff that matters outside of raw benches.
http://jsperf.com/ternary-vs-if-for-default-args