W3C home > Mailing lists > Public > public-html@w3.org > December 2013

RE: Proposed CR change to default bidi behavior for elements with dir

From: Erika Doyle Navara <Erika.Doyle@microsoft.com>
Date: Mon, 9 Dec 2013 23:54:53 +0000
To: "Aharon (Vladimir) Lanin" <aharon@google.com>
CC: "public-html@w3.org" <public-html@w3.org>, "member-i18n-core@w3.org" <member-i18n-core@w3.org>, Richard Ishida <ishida@w3.org>, Addison Phillips <addison@lab126.com>, "Sam Ruby (rubys@intertwingly.net) (rubys@intertwingly.net)" <rubys@intertwingly.net>, Paul Cotton <Paul.Cotton@microsoft.com>
Message-ID: <1cebf27ac37c4dec8ed02210b9b5f5dd@BN1PR03MB039.namprd03.prod.outlook.com>
Hi Aharon,

Do the subsequent changes [1][2] sufficiently address your concerns? If so, I will merge them into the HTML5.1 and CR branches of the W3C spec.


[1] https://github.com/w3c/html/commit/f397c78de13607ff21d27e112e4663447672a409

[2] https://github.com/w3c/html/commit/14f80c82877e0c61187d8af8aa1c9d4738294466

From: Aharon (Vladimir) Lanin [mailto:aharon@google.com]
Sent: Monday, December 2, 2013 12:41 AM
To: Erika Doyle Navara
Cc: public-html@w3.org; member-i18n-core@w3.org; Richard Ishida; Addison Phillips; Sam Ruby (rubys@intertwingly.net) (rubys@intertwingly.net); Paul Cotton
Subject: Re: Proposed CR change to default bidi behavior for elements with dir

Sorry I did not notice this sooner, but unfortunately, the changes in https://github.com/w3c/html/commit/cf3f3dbbc3a28e0b6f82d6395427c84b4bf489a8 (echoing the changes in http://html5.org/tools/web-apps-tracker?from=8282&to=8283) contain a number of serious problems, as described in https://www.w3.org/Bugs/Public/show_bug.cgi?id=23260#c20. I have suggested fixes there.

On Tue, Nov 26, 2013 at 2:41 AM, Erika Doyle Navara <Erika.Doyle@microsoft.com<mailto:Erika.Doyle@microsoft.com>> wrote:
As I've heard a number of statements of support (in person) and no objections, I have folded this change into the CR branch [1].

Please note: since the time I posted the initial spec change for review [2], Ian has produced a more comprehensive change [3] that updates some descriptions I had missed and that changes <bdo dir="ltr|rtl"> to be directionally isolated (to be unicode-bidi: isolate-override instead of unicode-bidi: bidi-override), as further requested by the i18n WG [4]. This keeps the behavior of dir consistent across different elements and enhances the usability of bdo under common usage. It is this later change (from the WHATWG spec) that I have chosen to fold into CR (as well as override my original change in the HTML5.1 master branch).

Please reply back with any comments or issues.


[1] https://github.com/w3c/html/commit/ca1986c79c26e49f7781577021b090df159c4c7a

[2] https://github.com/w3c/html/commit/efbf692e44fa3e6b9d66a0823fc2133ff1d18f1e

[3] https://github.com/w3c/html/commit/cf3f3dbbc3a28e0b6f82d6395427c84b4bf489a8

[4] https://www.w3.org/Bugs/Public/show_bug.cgi?id=23260#c17

-----Original Message-----
From: Erika Doyle Navara [mailto:Erika.Doyle@microsoft.com<mailto:Erika.Doyle@microsoft.com>]
Sent: Thursday, October 24, 2013 4:34 PM
To: 'public-html@w3.org<mailto:public-html@w3.org>'; member-i18n-core@w3.org<mailto:member-i18n-core@w3.org>; Richard Ishida; Addison Phillips; Sam Ruby (rubys@intertwingly.net<mailto:rubys@intertwingly.net>) (rubys@intertwingly.net<mailto:rubys@intertwingly.net>); Paul Cotton
Subject: Proposed CR change to default bidi behavior for elements with dir

The I18N working group has requested [1] that bug 22326 [2] be addressed in the HTML5 CR timeframe. This bug represents the implementation of the I18N proposal [3] to change the bidi semantics of the dir attribute to imply isolation rather than embedding.

The technical discussion so far has taken place on the WHATWG cloned bug [4], but here is a summary of the pro/con arguments (Richard/I18N members, please correct if I have missed or misrepresented anything):

Pro (make dir="" imply unicode-bidi:isolate) ==========================================
- HTML5 is consistent with  Unicode 6.3 spec (and other specs, like CSS3 Writing Modes) in making isolation the default for all inline bidi text embeddings

- Bidirectional text is easy and intuitive for web developers to implement correctly, and without needing to know anything about the intricacies of bidi algorithm

- Drives browser implementation to make the web less buggy with bidi issues. Problems with existing legacy pages will be fixed automatically.

- Easy transition period while browsers adopt new behavior (code will work better in browsers with the change, but work the same in browsers without)

- Once a browser supports isolation, it is trivial to implement this change in semantics for dir (simple change to default style sheet)

Con (dir="" continues to imply unicode-bidi:embed) ===========================================
- Currently no implementations yet of this behavior

Rebuttal: IE8-IE10 did a similar (but not identical) change, however, and there was little or no compat fallout, both according to IE bug records and also i18n community

- Default isolation for dir potentially makes the new HTML5 <bdi> element seem redundant.

Rebuttal: <bdi> element is still useful in cases when text of unknown direction is dynamically inserted to a page (because it has new HTML5 dir="auto" value as default, which determines dir at run time). Perhaps we could even re-label <bdi> from "bidi isolate" to "bidi insert".

...given the favorable reasons outweighing the unfavorable, I have implemented this change in the HTML5.1 Nightly draft, and believe it merits porting over to the HTML5 CR draft as well:


Please reply with any objections. Otherwise, I will port this change to the CR draft within a couple weeks (11/6).

Thank you,

[1] http://lists.w3.org/Archives/Public/public-html-admin/2013Oct/0015.html

[2] https://www.w3.org/Bugs/Public/show_bug.cgi?id=22326

[3] http://www.w3.org/International/wiki/Html-bidi-isolation#Proposal_A:_change_the_dir_semantics

[4] https://www.w3.org/Bugs/Public/show_bug.cgi?id=23260

Received on Monday, 9 December 2013 23:56:07 UTC

This archive was generated by hypermail 2.4.0 : Saturday, 9 October 2021 18:46:06 UTC