Re: [css-logical-properties] the 'inline-{start,end}' values for 'float' and 'clear'

2015-11-13 14:11 GMT-08:00 Johannes Wilm <johanneswilm@vivliostyle.com>:

>
>
> On Fri, Nov 13, 2015 at 11:03 PM, Koji Ishii <kojiishi@gmail.com> wrote:
>
>>
>>
>> 2015-11-13 13:59 GMT-08:00 Brad Kemper <brad.kemper@gmail.com>:
>>
>>>
>>>
>>> > On Nov 13, 2015, at 11:20 AM, Koji Ishii <kojiishi@gmail.com> wrote:
>>> >
>>> > Back up a bit, the original request was "is it safe to ship
>>> > 1-dimensional floats with logical directional values?" and while it'd
>>> > be ideal to figure out the final syntax for 2-dimensional floats to
>>> > answer to that question, I don't think it's absolutely required, as
>>> > long as we can agree that:
>>> >
>>> > 1. If no other 2-dimensional-related properties are set (e.g.,
>>> > float-reference), and
>>> > 2. If either 'start' or 'end' is specified
>>> >
>>> > we will handle it as 1-dimensional logical direction.
>>> >
>>> > I think this makes sense given the consistency with 1-dimensional
>>> > properties such a text-align, and shipping 1-dimensional logical
>>> > directional values before we finalize 2-dimensional syntax is
>>> > beneficial. Tab's response reads to me that Tab and fantasai are fine
>>> > with this.
>>>
>>> The problem is that it does depend on the final syntax somewhat. For the
>>> way I'd like it, start and end would be the values to use for one
>>> dimensional floats. But if 'start' ends up being short for 'start start'
>>> instead of 'start none', that that wouldn't work too well. It would change
>>> the meaning of 'float: start'. If we could agree that 'start' is short for
>>> 'start' in the inline direction and 'none' in the block direction, then I
>>> think 'start' and 'end' are the best choice.
>>>
>>
>> IIUC, it would change "only if float-reference is set to non-initial
>> value," so I still believe we can resolve 1-dimensional syntax
>> independently.
>>
>>
> Problem with that is that under Brad's proposal there would also be two
> dimensional inline floats which would have the same float-reference value
> as today's inline floats, as I understand it.
>

I'm not reading all 94 e-mails in this thread, sorry, so, ...ok.

We should avoid that though. Consider a case when we want to extend 1-d
property to 2-d, or 1-d/2-d impls are mixed among users. So just 'start'
should not have 2-d effects unless other properties or additional values
are set.

Could we try to reach consensus on this point?

/koji