GNU/Linux has two kernel sound drivers: OSS and ALSA. OSS is older and deprecated, being replaced by ALSA. Only one of these can be chosen as your sound driver.

On top of this ESD and ARTS are audio servers that accept requests from programs to play a sound, and then pass those sounds to the kernel sound drivers. Most Gnome applications use ESD, most KDE applications use ARTS, and some applications, like XMMS, can use either.

Traditional Linux kernel sound drivers are single-channel so that only one application can deliver audio to /dev/dsp at a time. Thus, if you were listening to music other sound events would not be heard. ESD and ARTS support multiple channels by multiplexing them and streaming the resulting audio to /dev/dsp. Some newer sound cards/chipsets support multiplexing in hardware and ALSA will make use of this.

Other functionality ESD and ARTS support includes streaming the audio to another machine.

Various commentary suggests that ESD has lower sound quality due to poor processing routines while ARTS has higher latency due to its more complex processing routines!