Re: RTL support in CSS (Was: [css3-box] Float start/end issue)

Le 24/09/2014 18:17, Julien Wajsberg a écrit :
> Le 24/09/2014 17:44, Tab Atkins Jr. a écrit :
>> On Wed, Sep 24, 2014 at 6:13 AM, Julien Wajsberg <jwajsberg@mozilla.com> wrote:
>>> Le 23/09/2014 22:22, Tab Atkins Jr. a écrit :
>>>> On Tue, Sep 23, 2014 at 8:29 AM, Julien Wajsberg <jwajsberg@mozilla.com> wrote:
>>>>> Tab Atkins Jr wrote:
>>>>> On Wed, May 28, 2014 at 2:17 AM, Salar Khalilzadeh <salar2k@gmail.com>
>>>>> wrote:
>>>>>> Back in 2009 a few values added to the float property in 'CSS basic box
>>>>>> model' module  but they haven't finalized yet.
>>>>>>
>>>>>> These new values for float property would be very helpful for RTL
>>>>>> languages.
>>>>>> Please give the module more attention.
>>>>>>
>>>>>> I wanted to hint on the Issue 61 which says "Adding ‘start’ and ‘end’ was
>>>>>> decided at 2009-12-02 telcon. Precise definitions not yet decided: does it
>>>>>> depend on ‘direction’ of the element itself or its parent? "
>>>>>> http://dev.w3.org/csswg/css-box/#the-float-property
>>>>>>
>>>>>> I strongly believe that start/end should depend on the element itself.
>>>>>> They
>>>>>> should follow the behavior of the float which applies on the element, so
>>>>>> the
>>>>>> new start/end values will depend on the direction of the element.
>>>>>> Otherwise I as a developer have to wrap my element with another element
>>>>>> and
>>>>>> change the direction there!! what a waste.
>>>>> On the other hand, that means you can't set "float: start;" on a bunch
>>>>> of elements in some container and expect them to float to the same
>>>>> side.
>>>>>
>>>>> We've addressed this in the Alignment module by having start/end base
>>>>> themselves off the container's direction, and having separate
>>>>> self-start/self-end values that base themselves on the item's
>>>>> direction.
>>>>>
>>>>>
>>>>> Is it something there is a good agreement about?
>>>>>
>>>>> I also agree that "start" and "end" should be relative to the containining
>>>>> box's direction (so, I don't agree with the initial mail), for consistency
>>>>> with other specs, especially [css-position-3], where we alway refer to the
>>>>> containing box.
>>>>>
>>>>> I have no opinion about self-start/self-end, but from the initial message in
>>>>> this thread it looks like there is at least some request.
>>>>>
>>>>> What's needed to move forward about this?
>>>>>
>>>>> [css-position-3] http://dev.w3.org/csswg/css-position-3
>>>> What do you think needs to be moved forward on?  Alignment is stable
>>>> in this regard; Box isn't a good reference for these topics (2.1 is
>>>> still the correct reference).
>>> Well, I'm looking forward seeing the "start" and "end" value specified
>>> and implemented for the "float" property. I don't see this discussed in
>>> 2.1 at all.
>>>
>>> Some context can be useful. I work on a core Firefox OS app and as I
>>> result I have the luxury to be able to use new CSS features, and give
>>> early feedback. These days we're working on supporting RTL languages.
>>> The usual way of doing this is overriding all properties that use
>>> left/right. There are tools that do this automatically. However while
>>> this is good for one-shot conversion, this is really bad for long-term
>>> maintainability. That's why I'm looking for the "native" CSS support for
>>> this. "start" and "end" values for the "float" property is only one part
>>> of the support obviously.
>> Ah, gotcha.  Unfortunately, there's no real timeline for continuing
>> work on 'float' right now. :/
>>
> Besides float, I'm interested in other RTL matters too:
> offset-start/end, support in background-position, margin-start/end,
> padding-start/end (and support in the shorthand notations). I think
> border properties are important as well, but not so important for me.

I forgot "transform": if we can find a way to make them direction-aware,
it would be nice. Maybe with a "transform-direction: obey-bidi/legacy"
property ?

-- 
Julien