'&' in Integration broker message

Question

We are currently sending information from a third party system to PeopleSoft financials using Integration Broker and I have run into somewhat of an issue. The message from the outgoing system has a field called 'DESCRIPTION'. One of our messages recently failed because the message contained the following value for the description 'substations & switch yards'.
(Colon will be used instead of semicolon as the terminator for the escape sequence to keep the browser from unescaping the escape sequence.....)
The description looks like : <DESCRIPTION>substations &amp: switchyards </DESCRIPTION>
As I would expect the '&' symbol is escaped in the output.
The problem is PeopleSoft does not seem to be handling this correctly. When the message arrives PeopleSoft turns it into the following:
<?xml version="1.0"?>
<AsyncRequest>
<data PsNonXml="Yes">
<![CDATA[.........<DESCRIPTION>substations & switchyards </DESCRIPTION>
]]>
</data>
</AsyncRequest>
It appears to unescape the &amp: appropriately but then does not treat the resulting xml as valid. It says it is non PS xml.
If in the stylesheet I say <xsl:value-of select="//DESCRIPTION" disable-output-escaping="yes"/>
Then the message is successfully processed, meaning it is recognized as valid PeopleSoft xml but the description is truncated to become
<DESCRIPTION>substations </DESCRIPTION>
It appears the parser cannot parse the '&' symbol if it is not escaped.
I was hoping someone could verify that this is in fact the case and suggest a workaround if it is. In my opinion it is not the behavior that would be desired. It should handle the parsing of the &amp: correctly and treat the resulting xml as valid PS xml.
In our case there is a premium on the number of characters that can be sent therefore doing something like translating the '&' to the word 'and' in the stylesheet is not the ideal solution. It would also be possible to disallow the '&' from being entered in the description field of the source system but this is not ideal either given the same constraint as above.
I believe somewhere I read a suggestion to move the data to an attribute but I don't think this is a viable solution either. I would assume since something like CopyRowset and CopytRowsetDelta "copies all data from the source rowset to the like-named fields in the message object executing the method" that it would not transfer the description from the message to the record if it were in an attribute as opposed to text as a child of the element. Assuming it would even be valid according to the xsd to begin with.
Ryan

Answer 1

Hi Jim,
If I capture the output generated by the third party system, transform it locally using the xsl the resulting xml is in fact well formed. Also as long as the description does not contain any special characters the result of the transformed message from the third party system is successfully processed by PS. This leads me to believe it is a direct a result of the special characters and not the format of the message after being transformed. Given that we are using Oracle's SOA Suite I am also able to capture the payload after it is transformed and prior to being routed to PS and this confirms the local test as well. The '&' is properly escaped and should be handled not as a special character but as simple character data.
The structured vs. unstructured is an interesting idea. Hadn't thought of it but will let you and the community how it goes if I go that route.
I may take your advice and go home for the weekend :)
Thanks for the insight.
Ryan