Vorbis dev - Oct 1999 - Streaming

I thought I might try experimenting with streaming.  There isn't anything
too odd I should be worried about, is there?

Tony Arcieri

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/

> I thought I might try experimenting with streaming.  There isn't
anything
> too odd I should be worried about, is there?
Jack and I have been talking about this for a few days off line...

The short answer is no, although there are two additional issues to think about
over the coding done so far.

First is chained streams; this is not something you need to worry about only
for 'streaming', but it needs to be added to the code in general.  In
the
context of Vorbis, this is a fancy way of saying that multiple
'individual'
vorbis streams can be concatenated into a single valid bitstream (so long as
the stream serial numbers are different). I've mentioned that before and
pledged example code :-)  This is only really an issue when you're seeking.

Second is datagram-based transmission, for example unicast or multicast UDP. If
we're talking about a real stream (eg, TCP), then you don't have much to
worry
about; the file representation is optimal for any lossless stream without
record boundaries.  If we talk about datagram-based broadcasts, the exact
structure we add to the datagram (over the raw audio packet) is a bit different
than what we want to use in a monolithic file.  So, if you're talking about
'internet radio' style streaming, we need to spec out how the data is
packetized first, work that isn't specced or coded.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/

I was thinking more along the lines of retrieving a stream of no
predetermined length via HTTP (using TCP as the transport protocol, of
course) as this is what xmms is most inclined to do.

Tony Arcieri

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/

> I was thinking more along the lines of retrieving a stream of no
> predetermined length via HTTP (using TCP as the transport protocol, of
> course) as this is what xmms is most inclined to do.
Yes, in this case, no worries.  Works just like regular decode.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/