On Tue, 13 Apr 2010 15:28:54 +0800, Jonas Sicking <[email protected]> wrote:
On Sun, Apr 11, 2010 at 11:39 PM, Silvia Pfeiffer
<[email protected]> wrote:
On Mon, Apr 12, 2010 at 4:00 PM, Philip Jägenstedt <[email protected]>
wrote:
On Mon, 12 Apr 2010 08:47:33 +0800, Silvia Pfeiffer
<[email protected]> wrote:
On Mon, Apr 12, 2010 at 7:59 AM, Jonas Sicking <[email protected]>
wrote:
On Sun, Apr 11, 2010 at 5:30 AM, Silvia Pfeiffer
<[email protected]> wrote:
f>> Is it expected that all of TTML will be required? The proposal
suggests
'starting with the simplest profile', being the transformation
profile.
Does
this mean only the transformation profile is needed to provide
subtitle
features equivalent to SRT?
That is also something that still has to be discussed further.
Initial
feedback from browser vendors was that the full TTML spec is too
complicated and too much to support from the start. Thus, the
implementation path with the TTML profiles is being suggested.
However, it is as yet unclear if there should be a native parsing
implementation of TTML implemented in browsers or simply a mapping
of
TTML markup to HTML/CSS/JavaScript. My gut feeling is that the
latter
would be easier, in particular since such a mapping has been started
already with Philippe's implementation, see
http://www.w3.org/2009/02/ThisIsCoffee.html . The mapping would need
to be documented.
Personally I'm concerned that if we start heading down the TTML path,
browsers are ultimately going to end up forced to implement the whole
thing. Useful parts as well as parts less so. We see this time and
again where if we implement part of a spec we end up forced to
implement the whole thing.
Things like test suites, blogging advocates, authoring tools, etc
often means that for marketing reasons we're forced to implement much
more than we'd like. And much more than is useful. This is why spec
writing is a big responseibility, every feature has a large cost and
means that implementors will be working on implementing that feature
instead of something else.
Understood. But what is actually the cost of implementing all of TTML?
The features in TTML all map onto existing Web technology, so all it
takes is a bit more parsing code over time. And if we chose not to
implement TTML, we will have to eventually support some other format
that provides formatting and positioning capabilities, seeing how the
legal landscape has evolved for traditional media (e.g. TV, set-top
box technology). Since TTML was originally developed to be the
exchange format for all such formats, it should have a sensible set of
features for this space. So, I personally think it's not a bad choice
for the purpose. Which other format did you have in mind to replace
it?
For the record, I am also not enthusiastic about TTML, specifically the
styling mechanism which even makes creative use of XML namespaces. An
example [1] for those that haven't seen it before:
<region xml:id="r1">
<style tts:extent="306px 114px"/>
<style tts:backgroundColor="red"/>
<style tts:color="white"/>
<style tts:displayAlign="after"/>
<style tts:padding="3px 40px"/>
</region>
...
<p region="r1" tts:backgroundColor="purple" tts:textAlign="center">
Twinkle, twinkle, little bat!<br/>
How <span tts:backgroundColor="green">I wonder</span> where you're at!
</p>
While I don't have any suggestions about what to use instead, I'd much
prefer something which just uses CSS with the same syntax we're all
used to.
[1]
http://www.w3.org/TR/ttaf1-dfxp/#style-attribute-backgroundColor-example-1
I have looked at alternative formats the provide styling and
positioning functionality. There is, for example, ASS/SAA
http://www.matroska.org/technical/specs/subtitles/ssa.html . We could
decide to support something like that instead, but it would be
essentially the same work as for TTML: define a mapping to Web
technologies (HTML, CSS, JavaScript) and implement that mapping, as I
don't think anyone would implement SSA natively either.
I am myself not excited by the way that TTML turned out and would have
wished for a more Web-friendly format to have come out of the W3C, but
that's what it is now. Also, I am really missing hyperlinking
functionality in it, but believe it is possible to extend it in the
future.
We could, of course, decide to develop a totally new specification,
but then we are left on our own to push that into the world. At least
TTML already has support by several existing vendors in the caption
space (see e.g.
http://www.adobe.com/accessibility/products/flash/captioning_tools.html,
http://www.ooyala.com/www3/support_closedcaptions,
http://broadcastengineering.com/automation/ninsight-unveils-dfxp-subtitling-mxf-ayoto-0901/).
I just think of all the options available TTML is the least
problematic way to go. But if somebody has a better idea, I'm more
than open to it! (This is me personally saying it - it's not a
representative opinion of the W3C HTML5 accessibility task force).
Henri Sivonen brought up an interesting point on the HTML WG list.
Bringing it up here as I don't know if people are following both
lists.
Will implementations want to do the rendering of the subtitles off the
main thread? I believe many browsers are, or are planning to, render
the actual video graphics using a separate thread. If that is correct,
do we want to support rendering of the subtitles on a separate thread
too?
Or is it enough to do the rendering on the main thread, but composit
using a separate thread?
If rendering is expected to happen on a separate thread, then CSS is
possibly not the right solution as most CSS engines are
main-thread-only today.
At least for us, the video (and audio) is decoded on another thread, but
composited on the main thread. Even if this were to change in some way, I
don't see that it makes any difference to captions. As long as it possible
to overlay arbitrary HTML on top of video without a performance loss
(which will be the case since scripted controls depend on it) then
rendering captions should be no more of a problem.
The remaining issue is sync, but I don't think compositing captions with
the video frames in the decoding pipeline (if that is what is being
implied) is the solution, because that would require deep(er) integration
with the media framework and a layout engine usable from a separate
thread. Neither are impossible, but don't seem to be justified just to
solve the sync problem.
--
Philip Jägenstedt
Core Developer
Opera Software
