Re: A new proposal for how to deal with text track cues

On Fri, Jun 14, 2013 at 6:13 PM, Silvia Pfeiffer
<silviapfeiffer1@gmail.com>wrote:

> Be honest to yourself: how much further is TTML? It doesn't have a
> rendering algorithm yet,


It depends on what you mean by "rendering algorithm". The TTWG believed
(and I think this hasn't changed) that defining a new rendering and
formatting system was both unnecessary and undesirable. That is why it was
formally defined in terms of XML-FO semantics, which is a well defined
model. As a point of convenience, we are now undertaking to define an
alternate model based on CSS semantics. In both cases, we do not wish to
define a new "algorithm" as such. A number of us look at the predisposition
to define algorithms in HTML5 as an exercise in over-specification. In
particular, there are other ways to increase interoperability, such as more
testing. In any case, nobody has proposed defining a more specific
algorithm to date.


> it doesn't have a JavaScript API that
> browsers would need to implement it properly (so application
> developers can manipulate the format),


Why do application developers have to manipulate its format in a client?
Such a use case is not a "given", and has not even been suggested for TTML.
Indeed, I know of no functional reason at this juncture that such a need
exists. I would be interested, though, to learn why you seem to assume the
opposite.


> it still has bugs that prevent
> interoperable implementations in details.


Do you know of any W3C specification that doesn't have bugs? This is not a
relevant comment. So far, only one very minor substantive change is
proposed for TTML1.0SE, the default value for ftp:markerMode. I appears to
me that TTML1.0 has proven to have very few bugs indeed.



> WebVTT even has a validator:
> http://quuz.org/webvtt/.
>

I can't speak for what internal tools are being used to validate TTML, and
at least one early implementation of a TTML presentation processor
(dfxpvw), performed a reasonable amount of validation, circa 2005. A new
TTML verifier tool (including validation) is now available at
https://github.com/skynav/ttv, and expected to be functionally complete
before the end of this month. I expect a web based interface to this tool
to come online soon as well.


>
> Please don't get me wrong: I'm not trying to attack TTML


Perhaps not, but some of your information is wrong or out of date.


> - HTML is
> another example of a format that continues to change. I'm just trying
> to put this statement into perspective. All formats always evolve to
> support more features. Bugs continue to be found and fixed. That
> doesn't mean that the format is a "half-finished strawman", which,
> incidentally, is rather derogatory language and doesn't sound very
> welcoming to anyone trying to work with WebVTT .
>

The fact that both the TTML community and the WebVTT community is either
incorrect out of date in its characterization of the other community
demonstrates that there is a need for better cooperation if for no other
reason to simply improve the common understanding of both communities.

Opposition to such cooperation seems to thwart such understanding and
serves only to codify misunderstanding.

Insisting that browsers will only implement support for one format is also
another barrier.

To reach a common ground, both sides will actually have to make an effort
to fully understand the technical details of both formats, and I'm not sure
I know of anyone yet who has this level of understanding (Dave Singer
perhaps?).