# given the following input doc:
<foo></foo>

# run through this stylesheet:
<xsl:stylesheet version="2.0" exclude-result-prefixes="#all"
    xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xmlns:A="urn:A">

    <xsl:output indent="yes"/>

    <xsl:template match="/">
        <bar xsi:type="A:AType">
            <xsl:value-of select="text()"/>
        </bar>
    </xsl:template>

</xsl:stylesheet>

# produces this result (Saxon-B, AltovaXML)
<bar xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:type="A:AType"/>

The problem with the result document is that it doesn't include the namespace declaration for the 'A' prefix, used in xsi:type. If I remove exclude-result-prefixes="#all" then it IS included.

Anyone had a similar issue before? exclude-result-prefixes must be used with care considering this behaviour.

Well the result is correct in terms of the namespace declarations and namespaces used for element and attribute names. I don't think attribute values are relevant to how the XSLT serializer works.

--
  Martin Honnen
  Microsoft MVP - XML

If the system knew exactly which namespaces were used in your result document, then exclude-result-prefixes wouldn't be necessary. It's provided because the system doesn't know. Obviously if you exclude result prefixes that shouldn't have been excluded, you must take the consequences.

Michael Kay
http://www.saxonica.com/
Author, XSLT 2.0 and XPath 2.0 Programmer's Reference