A mode of MP3
encoding that results in considerable gain in quality on most sources. It works by not encoding the left and right stereo
channels separately, but instead encoding their sum and difference separately (it's trivial
to recreate the right and left signals from sum and difference).
Why does this give better quality? because usually the right and left channel of a stereo signal are almost identical, so the difference will contain not very much information and can thus be compressed very efficiently, which leaves more space (and thus more quality) for the sum signal that contains most of the information. Of course, this means that the quality will drop when the right and left channel differ a lot at some point, but it will still be no worse than if you encoded them separately (unless the encoder did something stupid like dividing the bitrate between sum and difference signals in a constant ratio instead of adjusting it at need).
All in all, Joint stereo is a Good Thing, and you shouldn't use an encoder that can't do it.