RE: [bidi] Re: BIDI?

Ø  I agree with Mati that seeing "com.google.docs//:http" instead of "http://docs.google.com" in an RTL context is no good, and I agree with Mohamed that the best results would be to set the IRI's base level according to direction of the characters used in the domain.
We’ve actually received feedback from BIDI speakers that com.google.docs//:http (or com.microsoft.www//:http) is actually preferred by some users.
We’ve also had some feedback (from different governments) that http://RTL1.RTL2.RTL3 would sometimes be the preferred form.  (where RTL is supposed to represent a RTL field/label)
The actual preferences seem to be dependent partially on the user’s culture and partially on other life experiences.  (Eg: computer scientists are less likely to prefer the right to left ordering).  Other user correlations similar to Hebrew and Arabic mathematical preferences seem to exist.
So, we’ve been leaning toward the following thinking:
* always order the labels/fields either from left to right or right to left.
* pick the initial direction from the user environment (eg: English gets left to right fields, Arabic gets right to left fields).
* allow the user to override the direction in their preferences.
We haven’t thought as much about IRIs within a document, we’ve been thinking that if we need to understand user expectations for the address bar first.  However, I think in that case of IRIs within a document, the primary language of the document count as the “user environment”.
Note that we aren’t really giving any weight to existing behavior and were trying to understand user expectations from a clean slate because it’s trivial to get IE’s address bar, or other apps, to do obviously-wrong things (like put http in the visual middle of the string!)
Anyway, that kind of seems consistent with Mark’s document The actual preference of which direction the labels are arranged seem to be dependent partially on the user’s culture and partially on other experience.  (Eg: computer scientists are less likely to prefer that form).  Other correlations similar to Hebrew and Arabic mathematical preferences seem to exist.  Note that lots of the discussion has been around the “address bar” experience, and less about IRIs embedded in random text.
-Shawn