Support for native non-PCM hardware input/output (DSD, etc.)
This may be a giant waste of time but I've been thinking recently about biggest shortcomings of "digital versus analogue" volume control: analogue path is always source of uncorrectable and undetectable distortion while being overcomplicated (which is why it became obsolete) and digital attenuation of PCM output happens in hardware's native, integer format which effectively reduces number of used bits and "squeezes" the signal into what it's allowed to use. So even fancy native 32bit DACs are reduced to their 16-24 counterparts. And the problem is exacerbated by the fact that these chips are designed for 120-140db range while people tend to use around 60db + occasional blasting of headphones to 80-100db.
HOWEVER, most 32bit DACs, that are designed with native DSD in mind, have high-frequency output modulators and oversample the shit (x8-256) out of input PCM data. Or at least mine AK4490 does. Its DSD input, as seems to be the case for TI and ESS chips, has digital attenuation bypass. And DSD format itself, I suspect, does not suffer from loss of precision IF the volume is reduced by attenuating the amplitude on 100% signal. Meaning, if PipeWire could do its processing in PCM, as it does, but instead of passing data at the end to the driver as PCM it would put attenuation filter (amount controlled by real or fake volume control associated with the device), do PCM-DSD conversion (reverse of DoP) and pass that off to hardware then truly lossless digital volume control could be achieved.
Additionally, introduction of pre-output conversion could overcome JACK's design limitation of forcing DAC's sample rate to entire pipeline. PW could decouple hardware and software rates by doing conversion there which would allow using maximum of hardware's abilities while removing massive unnecessary strain from CPU (such as 48-192k in software and 768k in hardware) which should make DAC output modulator's job easier… or, at least, more mathematically correct.