Re: 2.1 Revisit proposals for activating/canceling bidi isolation (Was: Re: Implementation status of "Additional Requirements for Bidi in HTML")

First, an important note: it's bdi, not ubi. The HTML5 editor and WG
preferred bdi because the "bd" serves as a good mnemonic for
"bidirectional", which is a known term. The "ub" does not remind anyone of
anyuthing. Also, it is good that it is similar to bdo, so users will see
that they are related. I agreed with this logic. In any case, we alsways
said that the names are tentative, and the HTML WG will likely change them.

Regarding the element vs. attribute issue, while I continue to favor making
it an attribute, I do not see anything new in the arguments brought forth
above. They were already made in the bug comments, but failed to convince
the editor.

Now, for some detailed comments:

> - Should the output element, which is likely to be
> used mostly for numbers, really be isolated, with
> no way to override it in HTML?[5]

In the absence of bdi as an attribute, the solution that I would prefer to
that (if we want a solution at all) is simply removing the
unicode-bidi:isolate from <output>'s default stylesheet. If the isolation is
needed, the document can say <bdi><output ...>...</output></bdi>

> Suppose we put the second point into the default UA style sheet,
> [dir=auto] { unicode-bidi: isolation;}

Yikes! Good catch! The HTML5 editor forgot to do it. I have reopened bug
10808 <http://www.w3.org/Bugs/Public/show_bug.cgi?id=10808> requesting he do
it.

> but then we need an syntax to cancel bidi isolation.
> I am proposing we add the 'off' attribute to the ubi
> element with the following default UA style sheet
> (cancel the isolation of the children of the <ubi>
> element to 'embed'):
> ubi[off] > * {unicode-bidi: embed; }

There is no way you are going to get them to add an attribute for such an
esoteric need when the author can take care of it in CSS (just explicitly
put unicode-bidi:embed on the element with dir=auto, or on <output>). I also
don't like the idea of wrapping an element in an element in order to control
the properties of the (inner) element.

Aharon

On Wed, Dec 22, 2010 at 8:15 PM, Kang-Hao (Kenny) Lu <kennyluck@w3.org>wrote:

> Hi bidi folks,
>
> This is an attempt to revisit some issues around the prosed featured of the
> 'ubi' attributes. As far as I understand, the original global 'ubi'
> attribute was rejected because
>
> [[
> Generally we only add global attributes when you'd use the attribute on all
> kinds of elements, not just phrasing content.
> ]] (see [1])
>
> but I think in principle we need HTML syntax for everything that's needed
> from 'unicode-bidi' of the current CSS3 Writing Modes module[2]. The current
> proposal in HTML5 requires the user to use <ubi> or <output>[3] to activate
> bidi isolation, which has following issues raised by Aharon[4]:
>
> - Longer syntax (need a closing tag)
> - I want isolation by default for dir=auto elements, but I want to be able
> to suppress it with ubi=off, which I can't do with an element.
> - Should the output element, which is likely to be used mostly for numbers,
> really be isolated, with no way to override it in HTML?[5]
>
> [1] http://www.w3.org/Bugs/Public/show_bug.cgi?id=10807#c27
> [2] http://dev.w3.org/csswg/css3-writing-modes/
> [3]
> http://www.whatwg.org/specs/web-apps/current-work/multipage/rendering.html#punctuation-and-decorations
> [4] http://www.w3.org/Bugs/Public/show_bug.cgi?id=10807#c8
> [5]
> https://docs.google.com/a/google.com/document/d/17nb3wlYkIG9MNtL1mlXVPOVe4QwyApiHRYim3nTi5CE/edit?hl=en&pli=1&ndplr=1
>
> For the first point, Aharon mentioned an example like <a><ubi>BLAH
> BLAH</ubi></a> is too ugly and I concur with that. Suppose we put the second
> point into the default UA style sheet,
>
>      [dir=auto] { unicode-bidi: isolation;}
>
> but then we need an syntax to cancel bidi isolation. I am proposing we add
> the 'off' attribute to the ubi element with the following default UA style
> sheet (cancel the isolation of the children of the <ubi> element to
> 'embed'):
>
>     ubi[off] > * {unicode-bidi: embed; }
>
> Ex:
>
>    <ubi off><output>BLAH BLAH</output></ubi>
>
>   <ubi off><a dir="auto">BLAH BLAH</a></ubi>  ( <a>BLAH BLAH</a> should be
> used instead if you don't want to apply dir="auto" to it)
>
> This is really just a tweak of the original ubi="off" proposal.
>
> Pros
> 1. It provides a way to cancel bidi isolation on <output>
> 2. It reduces <a><ubi>BLAH BLAH</ubi></a> to <a dir="auto">BLAH BLAH</a>
> but there's side effect, you might need <a dir="auto"
> autodirmethod="inherit">BLAH BLAH</a> in the end. lol
>
> Cons
> 1. Using dir="auto" for the purpose of initializing bidi isolation is a
> hack. But I am guessing that some people would like to have
>
>        [dir=ltr], [dir=rtl] {unicode-bidi: isolation;}
>
> instead of the current 'embed', but this is a non-starter (cause it breaks
> backward-compact) and this hack might provide a workaround.
> 2. The fact that
>
>       <ubi off><ubi>BLAH BLAH</ubi></ubi>
>
> is not isolated is maybe unintuitive. Also, <ubi off> would not propagate
> to other descendants.
>
> Frankly, it seems that having a global ubi attribute would be more
> intuitive, but I have no idea what the consequences of introducing a new
> global attribute are. Perhaps an API would be needed?
>
> Some side comparison between these two systems:
> <ubi>BLAH BLAH</ubi>                       <span ubi>BLAH BLAH</span>
>
> <control>BLAH BLAH</control>          <control ubi>BLAH BLAH</control>
>
> <a><ubi>BLAH BLAH</ubi></a>        <a ubi>BLAH BLAH</a>
>
> <a dir="auto">BLAH BLAH</a>           <a dir="auto" ubi>BLAH BLAH</a>
>
> <ubi off><control>BLAH BLAH</control></ubi>    <control ubi="off">BLAH
> BLAH</control>
>
> Some are shorter, some are longer. If there's any statistics on the
> percentage of cases when the user wants "dir=auto" to activate bidi
> isolation, it could be helpful for obtaining a quantitative result.
>
> As far as I known, a goal of the HTML5 editor is to let the users use u, b,
> i and so on instead of span, so this is likely to be a reason why he doesn't
> like the right hand side, which has <span ubi>. However, <ubi> looks just
> too dummy in case 3 and 5.
>
> Perhaps this comparison can lead us to the right track, whatever it is.
>
> If it is concluded that the RHS is better, maybe we should suggest the
> editor to deprecate <bdo> in favor of a new 'bdo' attribute to make it
> consistent. A comparison pair is listed as follows
>
> <a><ubo dir="rtl">BLAH BLAH</ubo></a>     <a dir="rtl" ubo>BLAH BLAH</a>
>
> with the follwodefault UA stylesheet
>
> [dir] { unicode-bidi: embed; }
> [dir=auto] { unicode-bidi: isolation;}
> [ubo] { unicode-bidi: unicode-override; }
> [ubi] { unicode-bidi: isolate; }
> [ubi=off] { unicode-bidi: embed;}
>
>
> Cheers,
> Kenny
>
>