Just a semantic thing, but it is still prior art... it's just probably not prior art that, on its own, is enough to invalidate these claims. It could still be relevant for obviousness purposes.
I know I happily streamed .au music from the sounds.sdsu.edu FTP server into /dev/audio (late at night, of course) in that era. Since it was an FTP server, I suppose there was a rough sort of DRM (maximum connections allowed) and metadata (directories and filenames describing the artist and song name). Even though I had seen a good deal of MBONE broadcasts, it was quite a novelty to hear a song streamed on demand.
Fair enough, but given almost any thought at that time, we'd have thought of metadata (probably not DRM because we wouldn't have believed something so stupid was needed).
Once you've got a network, walkmans, a "mic" input, and a speaker, this stuff is obvious and that's the point.
The way we did it was pretty trivial: just piping into rsh from and to /dev/audio. However it was certainly streaming of music by anyone's definition.
[1] Making the admins of said network mighty pissed off in the process -- I believe the whole campus was on 10 Mbps at the time.