On 3/24/14 2:45 PM, Harald Alvestrand wrote: > On 03/24/2014 04:35 PM, Mandyam, Giridhar wrote: >> >> ØSorry, re-reading your post I see you said "avoid any UA-introduced >> processing delay". I hope you meant driver, or my answer makes no sense. >> >> Actually -- no, I did not mean driver. I don't expect the UA to be >> able to be able to detect any kind of post-processing in the driver, >> and a static table may not work in this case anyways as drivers can >> be dynamically installed or replaced. But your answer is also valid, >> just for a different situation. >> > > I don't understand what point is being discussed here - the API model > doesn't expose any distinction between the UA and the driver, so this > is (in my opinion) an internal implementation matter. The spec justifies getNativeSettings() by describing operating systems and native device APIs as being distinct from the User Agent, so I think we're at the right level. I agree with Harald that to the API, the UA and its OS drivers are one and the same. However, I wanted to assure Giri that I don't think UAs would voluntarily invalidate their own constraint model, and that the OS drivers are the reason for this method (except for echoCancelation perhaps). .: Jan-Ivar :.
Received on Monday, 24 March 2014 20:36:33 UTC