[Jack-Devel] Pulseaudio driver

Thomas Brand tom at trellis.ch
Tue Mar 26 21:28:12 CET 2019


On Tue, March 26, 2019 19:38, Chris Caudle wrote:
>> On 2019-03-26 13:53, Thomas Brand wrote:
>>
>>> making pulse a backend like alsa and others.
>
> That is a very bad idea for various reasons.  If you think it is super
> useful then feel free to develop such a backend, or pay a contract
> developer to do it for you, but the people who originally developed jackd
>  and the developers who currently maintain jackd as volunteers understand
>  the architecture of jackd and pulseaudio, and pulseaudio is a high
> latency server which explicitly states that it will sometimes drop audio
> samples. That is in no way what you want in an audio production
> environment. The current jack-source and jack-sink modules are the
> appropriate way to handle this situation, jackd is the low latency server
> and controls the audio hardware, pulse becomes a client of jackd.
>

Sure, don't get me wrong. Jack can do low-latency, synchronous processing
without loosing a sample and it should access the hardware as directly as
possible and run at the hw's clock.

The nice thing about jack is that its model is generic enough to a point
where it can also run as a high-latency client on-top of a potentially
lossy/inferior audio infrastructure, while keeping its intrinsic behavior
(synchronous, ...). The consequences of putting jack on-top of pulse can
be irrelevant for a given usecase.

As you pointed out, using jack-source, jack-sink modules for pulseaudio is
a setup known to be working. There is no immediate need for an
alternative.
We can think of scenarios outside of the existing, since its at no cost.
jack is flexible enough to adapt to almost any scenario, if needed.

Greetings
Thomas





More information about the Jackaudio mailing list