[etoys-dev] Etoys in Arab, is it possible ?

[Yoshiki Ohshima]

At Thu, 22 Oct 2009 10:58:01 +0200,
Bert Freudenberg wrote:
> 
> 
> On 22.10.2009, at 07:52, Andreas Raab wrote:
> 
> > Korakurider wrote:
> >> On Thu, Oct 22, 2009 at 1:26 PM, Andreas Raab <andreas.raab at gmx.de>  
> >> wrote:
> >>> Korakurider wrote:
> >>>>  Oh, I see. Thanks.
> >>>>  I also found screenshot of Arabic WinXP, that is interesting.
> >>>>
> >>>> http://blog.way2.jp/system/wordpress/wp-content/2009/02/arabic-windows.jpg
> >>>>
> >>>> So rewriting of all of UI pieces would be needed...
> >>> Mirror it. OLPCVirtualScreen can deal with it just fine using  
> >>> WarpBlt for
> >>> the RTL layout. No changes required; everything would stick to the  
> >>> right
> >>> edge; x would go left in viewers etc. It'd be a pretty simple  
> >>> thing to do.
> >>> The hardest part would probably be getting the text rendering to  
> >>> actually
> >>> produce LTR rendering for RTL text since you're going to do the  
> >>> mirroring on
> >>> your own.
> >>    But it means (x,y) coordinate system  will be also horizontally
> >> mirrored.  Is that valid thing?
> >
> > Mathematically, or culturally? Mathematically, yes (it's just a 180  
> > deg rotation around the y axis), culturally I don't know (but I  
> > would generally expect that increasing horizontal values are mapped  
> > in the way of the primary reading direction).
> 
> Etoys uses the mathematical convention, origin is in lower left. I  
> thought about the mirroring option before but thought it would be  
> unacceptable - all downloaded projects would be mirrored too, and  
> would have the coordinate origin in the lower right.

  I thought mirroring the whole screen is a joke.  It is not like
these countries are dominated by left-handers.  Positive x goes toward
right.

> Also, we will always have a mix of LTR with RTL text so we'd have to  
> double-mirror the LTR spans (see the Latin text in Arab screen shot  
> above).

  Yes, Pango and typical rendering engines handle it.

> OTOH even if imperfect it would still be better than not supporting  
> RTL locales at all. We'd need someone familiar with this to comment.

  Right, possibly only scriptors (and possibly viewers) can do it
(namely, scriptors interior is RTL and viewer clings from left... the
latter may or may not be important) that would be some improvement.

-- Yoshiki