W3C home > Mailing lists > Public > public-i18n-bidi@w3.org > July to September 2010

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

From: Adil Allawi <adil@diwan.com>
Date: Wed, 25 Aug 2010 00:37:21 +0300
Message-ID: <4C743B91.5080305@diwan.com>
To: Shachar Shemesh <shachar@shemesh.biz>
CC: Behdad Esfahbod <behdad@behdad.org>, "Aharon (Vladimir) Lanin" <aharon@google.com>, public-i18n-bidi@w3.org
  I did make a test implementation of dir=uba in Firefox and you can 
download this from here: http://www.diwan.com/w/Minefield.zip. And I 
list the issues I had here: 

 From my experience, I would agree with Aharon that dir=uba should not 
be implemented as it has been specified. The main issues for me (apart 
from what has been discussed) is that this is a feature in markup 
without a direct equivalent in CSS.

As I understand it, the way the Firefox code works (and other browser 
projects I have worked on) is to map HTML attributes that affect layout 
(e.g. dir=rtl) into CSS and layout on the CSS alone. However, dir=uba 
breaks this method. The implementation would not be easy and a special 
case only for dir=uba. I would also expect this can create performance 
issues as the layout could not be optimised on the CSS rules alone.

Also, it relies on different inheritance rules to CSS. Specifically: 
|dir=uba| sets |unicode-bidi| to plaintext isolate and |unicode-bidi| is 
a CSS attribute that does not inherit. But dir=uba is inherited in 
markup so it requires |unicode-bidi| to be inherited. This needs to be 
done explicitly in code and not to rely on normal CSS inheritance rules.

I think this is a lot of work for a feature that, I believe, has limited 
benefit and can be adequately described using JavaScript. Again, it 
falls down to the use-case. Unless there is an overwhelming need for 
this feature and I would not want to jeopardize the whole proposal for this.

Received on Tuesday, 24 August 2010 21:37:52 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:24:37 UTC