First off: any Enterprise drive will work for a NAS environment. The reason the REDs come into play is because they are some of the cheapest drives out there with a 3-year warranty. HD OEMs have changed warranty policies drastically over the past 12-18 months; WD was the first to use the easy color-coded "you get this warranty with this color" approach for consumers. Their MTBF is the same as the vast majority of just about every other drive in the WD lineup, and of other OEMs as well. The main difference is that TLER is hardcoded on in WD RED and above; you don't get this with Green or Blue drives. And while you can try to software-enable TLER on the cheaper (and lesser warranty drives), you may be shit-outta-luck when they start to drop off your array.
Secondly: It seems that people are seriously confusing the concept of streaming. DSD, or any type of file that matter, should be able to be read from any device for any reason from any device that can collect it over your LAN (and subsequently buffer it and play it). Whether you set the device up as a DLNA server/service (usually unnecessary) or as a SMB (sharing over your network, common with M$ and subsequently SAMBA on Linux offerings), once it is seen, as long as your infrastructure is capable of transferring the data - you are set. The NAS device is not and should not be expected to do the actual playback (transcoding->decoding->pre-amp, etc.). Now if your playback medium is not capable of pulling files from a basic home network - that is the player's problem, not the NAS!
Thirdly: A bitperfect rip of a DSD file (i.e. SACD->ISO->conversion) to, say, 88/24 resolution, still remains bitperfect and is nearly indistinguishable from the original. Is it ideal? No. But does it function well? Yes - especially if you do not have access to DSD core files or the playback software (or output capability from the player to the DAC). This is where DoP (DSD over PCM) comes in, but should more aptly be called DSD over USB (as most manufacturers are using USB as the interface). Basically: the device bitstreams the DSD file by piggybacking it through a PCM channel - no conversion is done, and the software is responsible for picking up on the DSD core file and running with it. There would be no SQ difference based on the source material, or file transfer (assuming your infrastructure can handle it), but you would almost certainly notice a difference from the playback software - and since this segment of the market is practically devoid of software devs, it may be quite some time before we see the WASAPI-standard for DSD playback.
Lastly: Mike, the 2001 "study" you mentioned was actually ripped off from a 1980 study! And it was the same guy that, in essence, said rock music was harmful to your health. Any double-blind tests done of Hi-Rez vs. DSD (especially recently) have resulted in one famous science'y word: 'inconclusive.'
Per ted_b at computeraudiophile: DoP standard means that a 24/176k PCM signal carries the native DSD...but it's still native DSD, not converted.
