Re: Suggestion for mail to HTML 5 WG about bidirectionality markup in HTML 5

Hi Felix, all,

Felix Sasaki wrote:
> Hi all,
> this is a draft mail for my action item
> Felix
> Dear HTML Working Group,
> the i18n Core Working Group has looked into the definition of the 
> "dir" attribute in HTML 5 at [1] and would like to make the following 
> comments.
> 1) We think that HTML 5, like HTML 4, should be able to render 
> bidirectional text
> without a style sheet. It would break backwards compatibility to remove
> the ability of a browser to do so without CSS. Therefore in our 
> opinion, HTML 5 has to
> describe the expected behavior 

Yes. Bidi behaviour should be a matter of HTML. Moreover, about the dir 
attribute, it is essential that HTML states clearly *what does the dir 
attribute do, and (also) what it doesn't do*. This is important for both 
HTML authoring and rendering :
- for browsers, to avoid browser's dependent rendering of bidi texts 
(e.g. like right align texts when dir="rtl")
- for HTML authors, to correctly (not more and not less) markup Bidi 

For example, we can state that the dir attribute :
-  make two different directional runs flow from left to right or from 
right to left
-  make the bullets go to the right of the RTL-script lists
- don't make RTL words be displayed from right to left. It is done 
automatically by the bidi algorithm
- don't right align text with dir="rtl" (though most browsers do)
- ...

> in at least the detail of HTML 4 

I think HTML4 is not very clear about this point. HTML4 goes into some 
details on the Unicode bidi algorithm (e.g. embedding levels, bdo...), 
but doen't tell too much on how to use the dir attribute.  For example 
did you know that most browsers don't agree at all on how to display 
tables with mixed rtl/ltr texts, with ot without the dir attribute? (see, Opera, Firefox, Safari have 
differences in rendering -- on MacOS.)
For discussion, I have started a FAQ 
(, on the use the dir attribute. Need 
to be completed. More ideas are welcome.

About HTML5, I think that it should at the very least, make a strong 
reference to (or pick up texts from)
"What you need to know about the bidi algorithm" [1] and/or "Tutorial: 
Creating (X)HTML Pages in Arabic & Hebrew" [2]. I hope the HTML WG will 
pay a close attention to the dir attribute and the Bidi isuues in general.

Best regards,


> rather than
> leave it up to the "presentation layer".
> 2) The section about the <bdo> element does *not* leave the expected 
> behavior completely up to the
> presentation layer - which is confusing. Content authors need to know 
> if they should use CSS, if CSS would override the specified
> behavior etc.
> 3) We propose that you add a note making clear that using the 
> bidirectionality markup provided is preferred, rather than attaching 
> CSS styling to arbitrary markup such as <p> etc.
> 4) You should consider allowing two new attribute
> values for the "dir" attribute: 'rlo' and 'lro' for dir. You do not 
> need to remove the bdo element,
> but the new values will allow content authors to proceed to a scenario 
> we described in the ITS 1.0 specification, see [2]. It will also 
> provide some additional power to the authors, since they will be able 
> to attach dir="lro" to a block element.
> 5) Note that we do not want to impose a requirement on implementations 
> of HTML 5 to implement CSS. You could just reference CSS and define a 
> default stylesheet fragment. This would just mean that an HTML 5 
> implementation has to make things behave as if it used this CSS 
> default stylesheet
> fragment.
> 6) HTML 4 made an exception in that it said that you don't have to 
> support bidi rendering unless you actually render Arabic or Hebrew. We 
> are not sure such an exception clause is still approprate, but it may 
> be a way to make the adoption of the above a bit easier.
> 7) Further information about bidirectionality markup can be found at 
> [3] and [4].
> Best regards,
> Felix Sasaki
> [1]
> [2]
> [3]
> [4]

Received on Monday, 18 February 2008 23:44:04 UTC