W3C home > Mailing lists > Public > www-archive@w3.org > September 2003

Re: Summary of SSML/I18N bidi issues/positions

From: Martin Duerst <duerst@w3.org>
Date: Fri, 19 Sep 2003 10:39:17 -0400
Message-Id: <>
To: "Daniel Burnett" <burnett@nuance.com>, "Richard Ishida" <ishida@w3.org>, <jim@larson-tech.com>, <luc.vantichelen@scansoft.com>
Cc: <scott.mcglashan@pipebeach.com>, <jerry.carter@scansoft.com>, <paolo.baggia@loquendo.com>, <jk@us.ibm.com>, <www-archive@w3.org>, <w3c-i18n-ig@w3.org>

Hello Daniel,

Very good summary. Just some comments:

At 22:43 03/09/18 -0700, Daniel Burnett wrote:

>Here is my attempt to summarize the issues and/or positions
>around the bidi topic discussed Thursday.  Please forgive me
>if I have misstated a position.
>Motivation for this discussion:
>For accessibility reasons, many W3C markup languages are
>a) required to ensure that authors *can* write documents using
>  the markup language in such a way as to *guarantee* that the
>  content of the document can be rendered visually, and
>b) encouraged to be designed in a way such that visual rendering
>  of *any* content marked up with that language is possible.
>Of course, W3C has a strong interest in ensuring that their markup
>languages work equally well for all human languages in order to
>avoid unnecessary language/cultural bias and to provide the
>broadest usability for the specifications.
>Combinations of text segments from different human languages
>  that are typically rendered visually in different directions
>  (right-to-left or left-to-right) are referred to as
>  bidirectional or bidi text.
>Martin's claims: (hope I got this right!)
>1. Current algorithms to visually render bidi text using only
>  a) the text itself and
>  b) indications of the language of each segment are imperfect.
>2. The addition of Unicode representations of text directionality
>  to the above is still insufficient to render such text perfectly.

   I guess you mean the bidi formatting codes provided by Unicode.
   The problem is not that they are not adequate for rendering
   bidi text ('perfectly'), e.g. for plain text; the problem is
   that they can interact very badly with markup, leading to
   things very similar to <i><b></i></b> crappy HTML examples.

>3. To ensure correct visual rendering of a marked-up document
>  for all human languages, explicit directionality indicators
>  must be provided as part of the markup (and hence be a part
>  of the markup language itself).
>4. Bidi controls in (X)HTML may provide this functionality for
>  specifications that need it.
>SSML group claims/statements:
>1. This is a general issue for many of W3C's specifications and
>  is in no way unique to SSML.


>2. Current solutions presented for this (e.g. XHTML elements/attributes)
>  are incomplete because they were not designed to work  for all
>  XML-based specifications.

They were not explicitly designed for all XML-based specifications,
but they would still work easily with an extremely broad range
of XML specifications.

>3. If important to W3C, this issue should be addressed by work
>  that explicitly incorporates feedback and requirements from
>  all specifications that may conceivably be expected to provide
>  this functionality now or in the near future.  It should *not*
>  be done as a one-off discussion between the SSML group and the
>  I18N group.
>4. The SSML group will be happy to incorporate the results of
>  any such comprehensive effort into its specifications when available.
>5. It is irresponsible to the speech industry and W3C to require
>  the development and/or implemention of non-comprehensive interim
>  solutions for bidirectionality or to force the progress of the
>  current specification to halt until the comprehensive effort
>  described is complete.

I do not think that the current XHTML solution is in any way
significant way non-comprehensive. I also do not believe that
any 'comprehensive effort' will lead to a proposal that is
significantly different from the current XHTML solution.

Regards,    Martin.
Received on Friday, 19 September 2003 10:50:26 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:42:31 UTC