🚀 go-pugleaf

I need a custom win32 component to display unicode text on both W9X 
and XP.
I've been working on a Uniscribe based component last week (using 
catch22.net tutorial and CSSamp from Microsoft SDK).
I've written from scratch a word wrapping algorithm (since CSSamp is 
buggy: it works only for LTR scripts)
Everything works fine but word wrapping on bidi RTL scripts (my 
algorithm actually works on simple LTR or RTL scripts: the problem is 
word wrapping on mixed RTL and LTR scripts).
Both TextOutW and Gdiplus::DrawString produce wrong output when word 
wrapping bidi RTL scripts.

I found a test page here:
http://people.w3.org/rishida/scripts/samples/wrapping.html

My component works fine on all LTR scripts and gives partially wrong 
results on the the arabic and hebrew scripts (actually it doesn't work 
when i need to wrap between two words of an embedded LTR script; 
everything works fine if embedded LTR scripts are one word long).

Uniscribe ScriptItemize breaks the arabic text  from the above test 
page into 9 items:
|ARABIC A|456|-|123|ARABIC B| (|global design|) |ARABIC C|
the Uniscribe bidi algorithm says that all uppercase items, <-> are 
RTL (numbers are LTR with RTL layout) while lowercase english  and the 
two brake items around and are LTR items

When word wrapping between "global" and "design" i get:
line 1: | global|) |ARABIC C|
line 2: |-|123|ARABIC B| (|design|
line 3: |ARABIC A||456|

I don't understand how to implement this logic in Uniscribe i.e.:
1) how to interpret valid break points using fSoftBreak and 
fWhiteSpace in SCRIPT_LOGATTR when this flags are set on LTR-RTL item 
boundaries
2) i've understand how to break an item to perform simple wrapping but 
i can't understand the reordering logic needed to get < global> and <) 
 > on the first line and < (> and <design> on the second (note that 
Explorer keeps the blank between global and design as the trailing 
blank - before global- on the first line!)

Is there any Unicode guru out there?
I need HELP!
Thanks in advance.

(I'm sorry for my latin english :-)

P.S.1 I don't understand why wrapping 456-123 on the hyphen in the 
arabic text behaves different from wrapping 123-456 in the hebrew 
text. On both 456 goes on new line and 123 stays over! Why?