Re: per-paragraph auto-direction, a.k.a. dir=uba

Interesting points. I tend to agree that if we see value in first-strong
being exactly as defined by te UBA, then we should go all the way and make
the default LTR, not the inherited direction.

Aharon

On Mon, Sep 27, 2010 at 10:15 PM, Amit Aronovitch <aronovitch@gmail.com>wrote:

>
> On Mon, Sep 27, 2010 at 8:31 PM, Aharon (Vladimir) Lanin <
> aharon@google.com> wrote:
>
>> Well, there are two separate issues here. One is that the UBA mangles AN
>> phone numbers pretty badly. This has no direct impact on the HTML/CSS
>> proposal. I have no problem sending that in, but would prefer to delay until
>> after the HTML5 bug submissions are done.
>>
>> The other is to get bidi@unicode.org feedback on the modified
>> first-strong estimation algorithm I am proposing, in the context of the
>> dir=auto feature (not as a proposal to modify the UBA).
>>
>> However, given that fantasai and Mati and Amit have already expressed
>> reservations about my proposal, and that no one has seconded it, I now
>> wonder whether it's  worth troubling them over it.
>>
>> Aharon
>>
>>
> Supposing you remove the special-casing of the weak-types, then there is
> still the question of what to do when no strong characters are found -
> return the inherited direction, or return LTR.
>
> The UBA algorithm is only called when there is no "higher level protocol",
> so it must return something in this "inconclusive" case. The default was
> chosen to be LTR.
> However, in our case, we have the option to return the containing element's
> directionality (at least, I would not consider this option more complicated
> than the UBA). This would be less disruptive if the element is empty, but
> has child elements.
> On the other hand, note that if you return LTR for this case as the UBA
> does, at least the examples you described (EN dates and negative numbers)
> would work fine.
>
>  Amit A.
>
>