Re: Roll-up captions in WebVTT

On Wed, Feb 29, 2012 at 6:28 PM, David Singer <singer@apple.com> wrote:

> > I don't understand the resistance to allowing UAs to render roll-ups on
> their own.  It would work much more reliably, and due to not needing
> special authoring work it would work on most files instead of almost none.
>  It would be much more flexible; UAs could, for example, vary the number of
> displayed lines and scrolling speed based on screen usage with the chosen
> font.  (I doubt that there's actually any demand for roll-ups, so it would
> probably never be implemented, but that's how it's supposed to work.)
>
>
> There isn't any; but the UA still needs to know which lines of text fall
> into the same grouping so that they make one text block that can/is 'move
> up' (jump or scroll).  It's that grouping that we're mostly discussing.
>

I suggested a mechanism to do that earlier in the discussion (unrelated to
CSS transitions), I think.

(To be clear, since this is a long thread--I'm not arguing that WebVTT
necessarily should support roll-ups at all, but if it does, then I think
that sort of mechanism is far preferable to authors doing it by hand with
transitions.)

I don't see anything 'brittle' about CSS transitions, by the way.
>

It's not CSS transitions themselves which are brittle.  It's the need to
align them to the results of WebVTT positioning, in section 3.5, which eg.
tries to keep cues from overlapping.  If a UA's positioning puts text in a
different position than the author's (for example, a different font causes
the "Adjust the positions of boxes ..." step to have a different result),
then a transition written for one result will be wrong in another.

-- 
Glenn Maynard