Streaming vs direct file access

[audio1]

Hello. I'm about to build a music server system, and I know I could save time (and money?) by buying an off-the-shelf system, but I want the satisfaction of self-built.

I built my own amp/pre-amp, and I'm reasonably proficient with compiling Linux applications and drivers, though I haven't yet attempted cross-compiling for ARM.

 

So to my question:

Can someone please comment on the technical/practical differences between the "streaming" network audio model, as used by many commercial devices, compared to direct file access over a network, as used by Music Player Daemon?

 

It seems to me that MPD's file-access method, whether using NFS or SMB, would be technically more straightforward and elegant compared to the somewhat complicated streaming method?

Is the popularity of streaming only because of the ability to have multiple clients, and multiple outputs?

 

[CharlyD]

As I understand it, MPD is a server application which accesses music content files, provides directory services (parsing based on metadata, etc.), decodes the content then streams PCM data to a rendering client. Is this correct?

 

If this is the case, isn't the file access method simply through the OS file system?

 

[audio1]

Ah, I may have confused the issue by mentioning MPD.

 

Sure, MPD's output can be configured to stream, but this is not its default mode. Normally MPD outputs audio at the "server" end. Networked clients control the playback, thus making them glorified remote control units.

 

So forget about MPD for the moment, and consider a conventional audio player application such as XMMS/Rhythmbox/Amarok/Aqualung;

my question relates to how these applications would access files over a network, versus how a rendering client accesses a network stream.

 

[CharlyD]

If the only visibility the rendering client has to the content is through the file system, wouldn't the entire content file need to be transferred to the client (downloaded) to access the metadata?