Re: [office] Bi-directional text.

[robert_weir@us.ibm.com]

I'm no expert in this area, but a few weeks
ago I had a discussion on Bidi with Jonathan Ben Avraham,  who is
involved with the Hebrew OpenOffice project (http://he.openoffice.org)

I came to the conversation with the view
that the Bidi algorithm in Unicode was sufficient.  Jonathan pointed
out some of the complexities of real-world use and why the Unicode mechanism
alone is not sufficient.  

To paraphrase our conversation, to support
Bidi the format needs to support base direction as well as aligment.  If
you try to guess from the character encoding alone, it fails, in areas
like:

1. Mixed characters in one paragraph, such
as Latin characters embedded in a RTL language. Especially in technical
documents, text like "SSL" or "HTTP" or "TCP"
are typically not translated or transliterated, so remain in Latin characters.
 So, an algorithm which looks at the charaters of the initial text
of the paragraph  may lead to the wrong layout in these mixed cases,
especially if the paragraph starts with Latin characters but then changes
to Hebrew.

2. Tables.  You might have a table with
only Latin content in an otherwise Hebrew document.  A Hebrew reader
would expect the row labels for the table to be on the right, not the left
as you might guess if you looked just at the characters.  HTML solves
this with a dir attribute.

3. Alignment.  In general this should
be settable independently of text direction. 

That said, I believe ODF already supports
what is required for these more complex scenarios.  Attached is a
document (created by Jonathan) which demonstrates the range of possibilites.
 I think our task is to ensure that the ODF specification is clear
on how these more complex use cases work.  



-Rob


Lars.Oppermann@Sun.COM wrote on 05/10/2006 02:46:01
AM:

> You can have LTR text in a RTL paragraph by using left-to-right-mark
and 
> right-to-left-mark or RTL/LTR-embedding unicode characters in the
text. 
> unicode offerers these for exactly that purpose.
> In OpenOffice 2.0, the insert-menu offers functions to insert those
> marks when working in a CTL environment. The unicode bi-di algorithm
> will when be applied accordingly.
> 
> I can see, how the same can be done with direction-attributes for
> span-elements. Nevertheless, the current solution works and I am not
> aware of any use-cases not covered by it. Do you know of any?
>

bidi_test.odt