[g-a-devel]Character encoding of description strings

[gnome-accessibility-devel bernard-hugueney org]

Bonjour,
I'm currently using at-spi & gnome-java-bridge (CVS) to track 
OpenOffice1.1 usage.
I'm having troubles with the characters encoding.
It seems that the top content of a text file is used as a description 
for this document, but not with the usual utf-8 encoding.

I'm writing xml descriptions, and a snippet might explain the problem 
:
...
<GUI role= "text" name= "Paragraph 49" description= "Paragraph: 
En-tête " id="0x80b2608" >
<STATES>EDITABLE	ENABLED	FOCUSABLE	MULTI_LINE	MULTISELECTABLE	SHOWING	
VISIBLE	</STATES>
<COMPONENT x="167" y="246" width="816" height="19" />
<EDITABLE_TEXT/>
<HYPERTEXT/>
<SELECTION/>
<TEXT nb_characters="7" carret_offset="1" >
En-tête
</TEXT>

...

As you can seen the description field of GUI seems to be in ISOLatin, 
while <TEXT> content is in the expected UTF-8.

Are my expectations of an UTF-8 description string wrong ?
If so, how amI supposed to know the encoding ?

Thank you very much in advance for any help.

Bernard