1. Acoustically silent is better.
Irrelevant, I can turn off all my fans from a front panel and it still runs at temps below 50C.
The idea that DACs and amplifiers are so easily isolated or immune from such issues was with CDs, and especially is with PCs, a myth. It can be done, but it's hard to do, and involves additional, expensive and undesirable complexity in the design of the DAC. As John Westlake of Audiolab recently said on this forum, he's not heard a DAC that's truly transport-indifferent.
Citation needed. Wasn't his post about SPDIF? As far as I know, "asynchronous USB" is clocked AND buffered by the DAC - it can (thanks to USB) prevent the computer from sending additional data until the buffer is normalized (filled up if below 50%, emptied a bit if above). The playback itself is then controlled by the DAC.
Complete galvanic isolation (ie. using an
optocoupler with additional circuit, powered by AC source used for the DAC itself) is therefore possible at rather cheap prices, it even fits the latencies guaranteed by the USB protocol.
Interference caused by the computer doesn't matter in that case at all, no
practical point in shielding components inside the computer case.
The goal is simply to make stuff better: it's crazy to deliberately introduce a contaminant to a system on the premise that you may be able to partially neutralise it downstream. Just don't introduce the contaminant. How can that be wrong?
I'm all for making stuff "better", whatever that might be. Shielding stuff to prevent RFI is generally a good thing and, as I said, a computer that is designed in such way is a nice engineering feat. It makes perfect sense to make those, as it is with ie. expensive Porshe car lines.
The resistance you're encountering here on this forum comes mainly from people, that want to "get from place A to place B" using a car. Basically any car will do the same job and for us - simple folk - using a Porshe just seems way too pointless.
Grounding the computer case (as done by the PSU) and shielding transports that lead to external DACs seems to be the easiest solution to get the same target effect as carefuly shielding
everything. Though shielding everything is of course "better" in terms of electromagnetic radiation.
Cables for USB, SPDIF, ethernet, SATA: either all differ - or none do.
That's an invalid statement if "differ" is defined as "differ in final sound quality".
The main resistance we encounter is on forums from people who haven't heard them, and who keep missing the point by referring to a computer's ability to do sums fast. It's alarming.
It's not alarming at all. People here (especially the analytic kind) just want a car that performs best at a given price point. Not that your cars are vastly overpriced, but people here seem to want cars that simply do the job, without being built as best as possible.
What you say here shows a fundamental ignorance on your part of the difference between synchronous and asynchronous data transfer.
As I said, and you ignored, the best place for your music storage is on a bog standard PC or NAS physically remote from your music client on an ethernet network sending asynchronous packet based data under TCP or UDP.
As a computer engineer, I can say that your statements are not entirely true as well.
TCP is a synchronous protocol (although current implementations tend to allow a bit of desynchronization during high latencies, for the sake of throughput) - it requires the other side to acknowledge each packet, so that no data is lost. This adds latency (needed retransmition) on unreliable links, however you're safe on Ethernet as long as your player uses a (small) buffer for playback.
UDP is asynchronous, damaged packets are simply dropped, without notifying the sender. The drop also happens when the link is saturated and the sending device has a faster hardware - the receiving end just doesn't have enough processing power. In general, UDP is better for audio/video streams on unreliable media as a missing stream piece is usually better (especially for video) than the latency added by TCP.
UDPlite was designed to amplify this effect. While UDP still uses checksums to at least differentiate between "correct" and "damaged" data within a packet, UDPlite doesn't use this checksum at all - it's built on the idea that damaged content it still better than no content, especially when the content is a digital representation of an analog signal, like an audio/video stream.
For the most part, using just TCP over LAN is enough. Using an NAS box has no practical latency/shielding advantage over stashing the HDD(s) into the computer directly, though it makes sense for pretty-much-diskless computers like SBT.