[theora] [OT-ish] WebM/Ogg VP8 streaming

[Gregory Maxwell]

On Sat, May 22, 2010 at 11:20 PM, Chris Double
<chris.double at double.co.nz> wrote:
> On Sun, May 23, 2010 at 3:04 PM, Christopher Blizzard
> <blizzard at mozilla.com> wrote:
>> I would suggest that we push to change that.  Live streaming is a pretty
>> great use case.
>
> I'm not so convinced that live streaming is such a great use case with
> HTML5 video as specified. There's no way for a browser to identify
> something as a live stream. When the browser buffers the stream falls
> behind and the user can't 'catch up' without restarting the stream
> manually. There have been so many complaints about this sort of thing
> with live streaming using <video> that I think it'd be better to
> either add stuff to the <video> spec to make live streaming support
> better, or live stream using a different protocol (and build support
> in the browsers).

It's very important to point out that the "alternatives" only sound
great here until you actually try to use them.

For example,  RTP has enormous problems with NAT traversal ... even
when people don't block it entirely.  It also provides no loss
handling facility ... if you don' mind the latency TCP loss recovery
is a pretty nice thing, especially if you're loosing around 1:10000
packets, as thats enough to make the output unhappy but not enough to
make retransmission much of a burden.

HTTP streaming sucks in a lot of ways, even more when you're short on
bandwidth... but you can make it go almost everywhere HTTP can go and
it requires very little software in order to support it. This is why
shoutcast / icecast continue to be so popular.

In terms of promoting open media— keeping solid support for easy to
use, best effort, basically-works-everywhere would do well for US.
MP4 has significant problems on the live streaming front.