Re: Specifying the audio buffer size

Perhaps the most practical thing to do is to notate this as a lower bound
on latency. The purpose is to allow less latency-sensitive applications to
express their tolerance for higher latency. It will be pretty tricky to
guarantee a latency upper bound when running a complex audio stack on
arbitrary hardware.

On Thu, May 14, 2015 at 3:32 PM Jan-Ivar Bruaroey <jib@mozilla.com> wrote:

>  I meant the latency attributed to the audio buffer only.
>
>
> .: Jan-Ivar :.
>
> On 5/14/15 12:51 PM, Harald Alvestrand wrote:
>
> On 05/14/2015 02:42 PM, Jan-Ivar Bruaroey wrote:
>
> Audio doesn't always go to the JS, so that seems like an unclear
> definition. Don't know about hardware variations.
>
> Wouldn't it be easier to define latency strictly by the buffer size, since
> that's primarily what this setting would control? Seems to be where we
> started.
>
>
> Buffer sizes depend on things like sample rates and encoding formats - 8
> kHz mu-law has completely different buffer sizes from 44.1 kHz L16. And if
> you buffer an encoded format for a variable-bitrate encoder, the
> relationship between buffer size and delay is impossible for the user to
> predict.
>
> I like seconds. It's what the user is usually looking for.
>
>
>