[Ogg a11y] Milestones 1&2 for video accessibility project

[Henri Sivonen]

On Nov 26, 2008, at 13:48, ogg.k.ogg.k at googlemail.com wrote:

>> Henri and I had an interesting conversation about the charset
>> specification in OggText and the sentence about "charset implies
>> directionality". It seems I have looked at the wrong documents for
>
> I read "charset" originally as "set of characters used by the text we
> are looking at", etc, a range of code points usually, but after  
> looking
> at the message headers you specify for, eg, language, I also see
> a "charset=utf-8", so I think I may have misunderstood this.

charset in the IETF sense means a character encoding--not a set of  
characters.

> Unicode defines directionality for *some* ranges, in two "strengths".
> This is what I use (through Pango). The directionality that can be set
> in a Kate stream is used as a weak default only.
> Now, a script may imply a default directionality too. Languages map in
> a fuzzy way to scripts, so you could get a default directionality  
> from the
> language, I think.
>
> So, directionality could be Unicode based if known from the text we're
> parsing, then default to either script or stream declared  
> directionality if
> not known.

Markup-like formats should provide functionality corresponding to the  
dir attribute in HTML. Non-markup-like text formats should use LEFT-TO- 
RIGHT MARK and RIGHT-TO-LEFT mark to give the dominant direction of a  
block of text.

I think text track formats in Ogg shouldn't specify their own  
directionality handling any further that HTML does and should defer to http://unicode.org/reports/tr9/ 
  for the rest. Whether it's necessary to have exactly as much format- 
level directionality control as HTML, I'm not so sure about. (HTML has  
markup for declaring the dominant direction of a block (dir='...' on  
anything but <bdo>) and markup for overriding the intrinsic  
directionality of characters (<bdo dir='...'>).)

> Could there be cases where one might want to override the  
> directionality (eg, render English r2l or Arabic l2r ?)

I take it that you mean overriding the intrinsic directionality of  
characters. HTML, CSS and W3C TimedText provide this, but *I* can't  
think of legitimate captioning use cases off the top of my head.

-- 
Henri Sivonen
hsivonen at iki.fi
http://hsivonen.iki.fi/