Re: BIDI?

I have thought for a while about the bidi requirements for IRIs.

*Firstly*, I agree with Mark's general proposal - to have a uniform 
ordering of fields in IRIs. I have done some experiments with making 
Arabic tweets order right to left and the process is very similar. 
Figure out the 'fields' of rtl parts (e.g. Arabic phrases) and ltr parts 
(e.g. urls, smilies, #tags) then make sure these order uniformly 
right-to-left by inserting rlm and lrm marks.

The big problem is that Bidi IRIs are already a reality and have been 
since ICANN allowed local language TLDs in 2009. So our ability to mange 
user expectations is limited by how Arabic IRIs behave now. Here is a 
good example from Qatar:

http://الاعلي-للاتصالات.قطر/ar/news-events/event

or using capitals to replace the Arabic words:

http://QATAR.TELECOMS-SUPREME/ar/news-events/event

The url is generally left-to-right except for the domain which is 
right-to-left. Arabic users now expect the domain part of an IRI to be 
ordered RTL if it is in Arabic.

There is a need to have two directions for IRIs. I would suggest that 
the ordering direction is controlled by the direction of the language of 
the TLD. So السعودية domains should always draw rtl, but .sa will draw ltr.


*Secondly*, The text editing interface for a Bidi IRI is a nightmare for 
an Arabic user. There is no consistent cursor or highlight behaviour 
across browsers. There is also no consistent user expectation - does the 
highlight extend visually or logically across the text? How would this 
be entered on a mobile phone with a numeric keypad?

If one inserts an English letter inside the RTL domain, the whole domain 
will flip in around the English letter. How would a user understand 
about how to correct this?

To get around these problems I would like to see a subset of IRIs that 
can be entirely ordered RTL. That would include the scheme and the path. 
Like so:

ويب://الاعلي-للاتصالات قطر/ع/الصفحة_الرئيسية

There would need to be Arabic translations of some (not all) scheme 
names (e.g. http and ftp). This would allow web developers to create 
IRIs that are easily typed, edited, moved between applications and 
transferred unambiguously from paper to computer and back again. This 
could be as part of a transitional phase either:
- these are the only type of RTL IRI allowed until a large number of 
clients support a UBA extension.
- this is a voluntary IRI restriction that can be recommended and 
validated independently.

Adil

On 27/04/2011 00:22, Mark Davis ☕ wrote:
> Here are some rough thoughts on how we could handle bidi IRIs.
>
> http://goo.gl/QwSoo
>
> Feedback is welcome.
>
> Mark/
> /
> On Wed, Apr 20, 2011 at 23:20, Shawn Steele 
> <Shawn.Steele@microsoft.com <mailto:Shawn.Steele@microsoft.com>> wrote:
>
>     I'm wondering what the current thinking around BIDI IRIs is?  A
>     few things in draft-ietf-iri-3987bis-05 jump out at me.
>
>     -Shawn
>
>      
>
>     http://blogs.msdn.com/shawnste
>
>