[foms] Proposal: adaptive streaming using open codecs

Andy Berkheimer andyberkheimer at youtube.com
Mon Nov 1 18:08:41 PDT 2010

On Mon, Nov 1, 2010 at 6:33 PM, Christopher Blizzard
<blizzard at mozilla.com> wrote:
> On 10/29/2010 6:09 PM, Silvia Pfeiffer wrote:
>> Actually, my son frequently does that with the YouTube videos he
>> watches - he will open 20 tabs of YouTube videos put them al briefly
>> on "play" then "pause" so that they start downloading - then go off do
>> something else for a bit and come back later to watch one after the
>> other. It's very useful here in Australia with the delays and
>> fluctuations in bandwidth that we get. Also, it's on a desktop and he
>> has no storage problems.
> This is an anti-pattern to a lot of video providers.  I've talked with a
> few people now who use flash media streaming because they don't want
> buffering.  Lots of video providers have to pay for bandwidth (not all
> of them) and lots report people buffering 20 mins of video and end up
> only watching 1 minute of it.

I'd say the key as a provider is that I want pretty good control over
this.  The ideal is for just enough readhead to avoid stalling the
decoder pipeline.

[ Incremental network demand does have a marginal cost for YT and we
are sensitive to this too, contrary to some reports - just to clear
that up! ]

But there are also some scenarios where pre-buffering more data for
improved performance can be worthwhile, even if those bits are less
likely to be watched.  The two examples that come to mind: some users
prefer a consistent quality experience and are willing to wait for it,
and some users are in environments that are too slow to ever get a
stream in real time.


More information about the foms mailing list