On Mar 1, 2012, at 6:44 , Glenn Maynard wrote:
> On Thu, Mar 1, 2012 at 1:33 AM, Silvia Pfeiffer <silviapfeiffer1@gmail.com> wrote:
> I personally think it makes sense to drop such cues on the floor,
> since they should not have any influence on anything, never being
> active.
>
> They can be looked up by id, which can affect scripts, especially for metadata tracks.
>
> Perhaps they'd still be read by text-to-speech systems (eg. the "read on-screen text aloud" category of use cases), which I imagine would normally ignore end times entirely.
I hope they don't ignore end-times *entirely*; yes, there is an issue if TTS needs more time than the display time, but cues with zero or negative duration should be dropped and/or flagged as an error, IMHO.
(This is one reason why many systems use a start-and-duration model.)
> (These aren't use cases, just situations where dropping them on the floor would cause a detectable change in behavior, and that might be perfectly OK.)
David Singer
Multimedia and Software Standards, Apple Inc.