PulseAudio

PulseAudio is a background process sound server that attempts to intercept any software or hardware source of sound and direct it to sinks (destinations that receive sounds) as intended by the user. Any legacy or alternative systems such as OSS, ALSA, ESD, GStreamer, Xine, SDL, or aRts are to be intercepted even if they are aware of each other and try to play nice. In some cases this is especially necessary for proper sound. For instance, ALSA normally intercepts sound to emulate such things as OSS; which means multiple redundant layers of emulation would be slowing down the CPU if PulseAudio is not able to intercept OSS first. If the audio was meant for some other system, PulseAudio generally does not handle the audio itself but direct the process. Sound sent to ALSA goes to PulseAudio where PulseAudio sends it back to ALSA to be sent to the sound card. Obviously, it would be more efficient to send sound in a format understood by PulseAudio. But, when users attempt to run multiple sound systems simultaneously, PulseAudio is the only system that can successfully coordinate. The older systems typically surrender all audio processing to one application at a time. PulseAudio also has the advantage of being able to bidirectionally communicate over a network to any audio system and other PulseAudio servers.

PulseAudio works with JACK by becoming a client application to JACK.

OSS does all sound mixing in the Linux Kernel which is a violation of Linux Kernel design policy. PulseAudio does not, even when handling OSS audio.

PulseAudio is not meant to be "a single audio system" to replace all others. Quite the opposite. PulseAudio exists because there are many audio systems that are require to work together.

PulseAudio can:
Adjust audio for each application independently. For instance, mute while still recording one steaming audio from the Internet while the user listens to another.
Use plugins and loadable modules for anything the user would rather not use PulseAudio for.
Produce sound on modern systems for applications that use obsolete sound systems or hardware.
Measure and correct for latency without causing additional latency
Discover other sound servers and send sound to them over the Internet. Play audio from computer A on speakers of Computer B without requiring additional CPU use of Computer B.
Play any source to any hardware without the application needing to know how. Simultaneously play a Megadrive game and an SNES game on separate earphones. Play modern app sounds on legacy sound cards.
Run operations from command-line interface and script them.
Reconfigure any audio, including hardware devices, from the command-line without interrupting sound.
Convert samples and resample without interrupting sound.
Split a sound card into multiple virtual cards
Join multiple sound cards into one virtual card
synchronize multiple playback (even across networks)
Bluetooth audio works the same as any other audio.
Run on multiple operating system and do everything it does operating systems. (OSS and ALSA are Linux only)