Xalan problem with xsl:attribute and Japanese characters

[pradip sonar]

Hello,

I am having a problem with Xalan transforming a document containing Japanese
characters. (An initial note: the problem does not occur on my Windows
development machine, only on the Solaris test/production environments.)

The basic problem is that for any place I'm selecting XML node values to use
within an xsl:attribute block, if the text is double-byte it is converted to
the NCR representation instead of the actual characters. Under Windows, the
"real" characters appear within the attribute tag. I'm positive the source
XML is UTF-8 encoded, as I can output the value correctly within the same
stylesheet as long as it is not output within an xsl:attribute tag.

Here's an XSL snippet--the basic idea is I'm creating a bunch of HTML fields
and setting the "value" attribute to a selected XML element value:

<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output encoding="UTF-8" omit-xml-declaration="yes" method="html"/>
.........

<xsl:template match="//root/row">

<tr>
THIS VALUE IS KANJI AND GETS OUPUT AS NCR

<td width="400" class="label" height="18">
<input type="text" name="companyAffiliation" size="50">
<xsl:attribute name="value">
<xsl:value-of select="companyAffiliation"/>
</xsl:attribute>
</input>
</td>
</tr>

<tr>
USING A TEXTAREA WITHOUT XSL:ATTRIBUTE, IT OUTPUTS CORRECTLY

<td width="400" class="label" height="18">
<textarea name="companyAffiliation2" rows="1" cols="50">
<xsl:value-of select="companyAffiliation"/>
</textarea>
</td>
</tr>

</xsl:template>

Any ideas?