DSD is not new. DSD64 (1bit @ 2.8 MHz ) is known basically as SACD. This resolution is most widespread, and all downloads I am aware of are 2.8 MHz.
I am a free lance recording engineer. I always strive to make recording as close to the impression the listener in the audience in the best seat of any given venue would perceive live.
I have recorded, always pure 2 channel, analog casette without noise reduction, analog casette with noise reduction, S-VHS video casette, CD-R and recently DSD.
Korg DSD recorders I now work with besides DSD at 2.8 and 5.6 MHz ( DSD128) also support any PCM between 192/24 down to 44.1/16. I find DSD at 5.6 MHz better than anything else. It simply sounds closer to the mike feed than PCM even at 192/24.
Previous posters have presented pros and contras in almost every possible way. I am the first to admit DS is not perfect - but it most closely resembles what we can hear. It may have problems with noise above audio range, it may not be totally distortion free - but it is fast enough, it will record a pulse better than any PCM in existance, even DXD which is PCM at 384 kHz sampling frequency/32 bit depth.. Although I respect Mr. Lip****z very much, I feel that total mathematical perfection is less important than simply registering something has happened. An analogy: if you were some minority and your representative in Congress would have two options: because your last name/tribe/whatever is very hard to pronounce ( high frequency in audio, above 20 kHz ), in PCM not recorded at all ( but everything below that correctly, which does exclude you ) or in DSD slightly distorted ( mispelling your name ) - what would you prefer, not being mentioned at all or at least having a chance to clarify matters later?
SACD failed because vast majority of recordings, done conventinally, much the same as for CD, simply do not contain recorded information that could take advantage of DSD/SACD. Very few recordings actually originated as DSD recordings - "upsampled" PCM is not or only marginally better than in native enviroment - and majority of DSD recordings have been mastered in PCM environment and only at last stage converted back to DSD/SACD.
This was/is because hardware and software for mastering DSD in its native enviroment was/still is very expensive and only recently really pure DSD recordings start to become available.
DSD128 or DSD at 5.6 MHz is the first digital format I consider to digitize my vinyl library - it is also the first good enough to make recordings of actual phono cartriges tracking high frequency information from various test records. The best phono cartridges exceed capabilities of any present digital format - there were carts capable of 120 kHz bandwidth, and only DSD will record (attenuated ) so high frequency - any PCM ( with the possible exception of DXD ) will be long gone at this frequency. If you observe the 1 kHz square wave recorded on test LP on an oscilloscope, DSD at 5.6 MHz looks to naked eye as the best approximation of the output from cart/pream directly connected to scope - 192/24 can no longer capture fine detail in high frequency range ( due to for PCM characteristic ringing, which is MUCH better with DSD ).
Goodvibes has correctly indicated that analog stages may well be the limitting factor - digital, be it PCM or DSD, is perfectly capable of outstripping analog stages, which, if done right, are $$$ and are therefore prone to be the first victim of cost cutting measures when going from prototype to actual commercially available product.
It is expected that designers who have attained pinnacle in PCM will not be pleased of having to do it all over again - besides pure commercial considerations, which would of course tend to prolong present PCM as long as possible.
One of advantages of DSD is that it can be converted to any current PCM required - and you still have full resolution DSD master. If master is in PCM, that is it - resolution is as what it is and "upsampling" after the fact will do little if any good.
It is my firm belief that efforts the musicians have put into their performance should be recorded with the best possible "technicalities" - capturing as much of actual sound as possible - not limitting that due to current belief what is considered as audible or what hardware/software has reached general
Information missed at the recording can not be retrieved later on.