Icecast dev - Feb 2005 - Icecast2 Burst-on-Connect

On Thu, 17 Feb 2005, Michael Smith wrote:
> There's one thing that icecast doesn't do (except for ogg streams)
> that other servers might (but from what I'm told, shoutcast
doesn't):
> start sending from some valid synchronisation point (e.g. the mpeg
> sync pattern) - obviously, if you do that, it's a bit easier on
> clients (but  I don't know of any that don't cope).
Media Player Classic requires a Content-Length: before it will stream.

So does Windows Media Player 6.4, except it also requires something else - 
could it be this mpeg sync pattern ?

Now discuss "Content-Length: -1" ...

On Thu, 17 Feb 2005 11:22:49 +0000 (UTC), gARetH baBB
<hick.icecast@gink.org> wrote:
> On Thu, 17 Feb 2005, Michael Smith wrote:
> 
> > There's one thing that icecast doesn't do (except for ogg
streams)
> > that other servers might (but from what I'm told, shoutcast
doesn't):
> > start sending from some valid synchronisation point (e.g. the mpeg
> > sync pattern) - obviously, if you do that, it's a bit easier on
> > clients (but  I don't know of any that don't cope).
> 
> Media Player Classic requires a Content-Length: before it will stream.
We already do this (with some stupid faked value) for... the flash
client, I think. We could add another case for this thing - can you
give us details (user-agent string?).

Does it also work with HTTP/1.1 Transfer-Encoding: chunked? That'd be
a lot more work to implement, but it's the right thing to do in the
longer term.
> 
> So does Windows Media Player 6.4, except it also requires something else -
> could it be this mpeg sync pattern ?
> 
> Now discuss "Content-Length: -1" ...
?? What about it. Does this make some media players work, somehow?

Mike

On Thu, 17 Feb 2005, Michael Smith wrote:

[please don't post to the list and email

plus your Cc posts don't actually seem to make it to the list anyway]
> > Media Player Classic requires a Content-Length: before it will stream.
> 
> We already do this (with some stupid faked value) for... the flash 
> client, I think. We could add another case for this thing - can you give 
> us details (user-agent string?).
"Media Player Classic"

Unfortunately Windows Media Player 6.4 uses the IE User-Agent, presumably 
like all WMPs.
> Does it also work with HTTP/1.1 Transfer-Encoding: chunked? That'd be
> a lot more work to implement, but it's the right thing to do in the
> longer term.
Don't know.
> > Now discuss "Content-Length: -1" ...
> 
> ?? What about it. Does this make some media players work, somehow?
If you haven't got a fixed Content-Length then the alternatives are a made 
up *big* Content-Length (which is at least "valid") or a bogus 
Content-Length, such as "-1".

http://www.dabbar.co.uk/ chucks out a Content-Length of -1.

MPC and WMP6 both work with DABBAR perfectly.

But that isn't the whole story for WMP6, as I can't get it to work with
my
own MP3 Icecast2 streams and a Content-Length of -1, but if I use a 
<relay> in Icecast2 pointing at my DABBAR box then it will actually work 
ok with WMP6 - so it must be something part of the stream.

Which is why I was wondering about mpeg sync patterns. DABBAR as it's just 
chucking out a demultiplexed DAB stream is admittedly chucking out MP2 and 
not MP3.