You may not need to use CDATA at all. XML GENERATE will take the content of XMLFILEINH and escape the special characters (as you have indicated). The resultant XML when viewed with a simple text editor will show escape sequences - not what you want. However, when you use XML PARSE to process it, the escaped characters will again be replaced with their original contents. Also, most XML aware viewers (e.g. Microsoft Edge among others) will display content as you expect without the escape sequences.
Here is an example IBM Enterprise COBOL 6.2 program illustrating my point:
IDENTIFICATION DIVISION.
PROGRAM-ID. XML5.
DATA DIVISION.
WORKING-STORAGE SECTION.
01 REQUEST.
06 ROUTE.
11 NAME PIC X(030).
11 VERSION PIC 9(004).
06 QUESTION.
11 IDENT PIC 9(009).
11 XMLFILENAME PIC X(006).
11 XMLFILEINH PIC X(5000).
01 XML-DOC PIC X(5000).
01 XML-IDX PIC S9(9) BINARY.
01 XML-CHAR-CNT PIC S9(9) BINARY.
PROCEDURE DIVISION.
MAINLINE SECTION.
MOVE 'serviceRequest' TO NAME
MOVE 1 TO VERSION
MOVE 111111111 TO IDENT
MOVE 'FILE-1' TO XMLFILENAME
MOVE '<?xml version="1.0" encoding="UTF-8"?><SOAP-ENV:Envelop
- 'e.....<SOAP-ENV:Envelope>'
TO XMLFILEINH
INITIALIZE XML-DOC
XML GENERATE XML-DOC FROM REQUEST COUNT IN XML-CHAR-CNT
PERFORM VARYING XML-IDX FROM 1 BY 80
UNTIL XML-IDX > XML-CHAR-CNT
DISPLAY XML-DOC (XML-IDX : 80)
END-PERFORM
XML PARSE XML-DOC PROCESSING PROCEDURE XML-HANDLER
ON EXCEPTION
DISPLAY 'XML Error: ' XML-CODE
GOBACK
NOT ON EXCEPTION
DISPLAY 'ALL DONE.'
END-XML
GOBACK
.
XML-HANDLER.
DISPLAY XML-EVENT (1:22) ':' XML-TEXT
.
The output is:
<REQUEST><ROUTE><NAME>serviceRequest</NAME><VERSION>1</VERSION></ROUTE><QUESTION
><IDENT>111111111</IDENT><_XMLFILENAME>FILE-1</_XMLFILENAME><_XMLFILEINH><?xm
l version="1.0" encoding="UTF-8"?><SOAP-ENV:Envelope..
...<SOAP-ENV:Envelope></_XMLFILEINH></QUESTION></REQUEST>
START-OF-DOCUMENT :
START-OF-ELEMENT :REQUEST
START-OF-ELEMENT :ROUTE
START-OF-ELEMENT :NAME
CONTENT-CHARACTERS :serviceRequest
END-OF-ELEMENT :NAME
START-OF-ELEMENT :VERSION
CONTENT-CHARACTERS :1
END-OF-ELEMENT :VERSION
END-OF-ELEMENT :ROUTE
START-OF-ELEMENT :QUESTION
START-OF-ELEMENT :IDENT
CONTENT-CHARACTERS :111111111
END-OF-ELEMENT :IDENT
START-OF-ELEMENT :_XMLFILENAME
CONTENT-CHARACTERS :FILE-1
END-OF-ELEMENT :_XMLFILENAME
START-OF-ELEMENT :_XMLFILEINH
CONTENT-CHARACTERS :<?xml version="1.0" encoding="UTF-8"?><SOAP-ENV:Envelope.....<SOAP-ENV:Envelope>
END-OF-ELEMENT :_XMLFILEINH
END-OF-ELEMENT :QUESTION
END-OF-ELEMENT :REQUEST
END-OF-DOCUMENT :
ALL DONE.
Note escaping of special characters in the "raw" dump of the generated XML, but upon completion of XML PARSE they are restored to what was given to XML GENERATE. This is normal XML processing. Character escaping such as this may protect you from code-page conversions when transmitting the generated XML. When using CDATA there is a possibility of corruption when the document has to be converted from one code page to another and there is no direct mapping for a given character (not likely but possible).
What I find interesting here, and can't explain, is why the generated XML tag names beginning with XML are prefixed with an underscore.
Final note: If the content of COBOL variable XMLFILEINH contained the sequence </_XMLFILEINH> somewhere one might think that it would cause premature ending of the <_XMLFILEINH> tag in the resulting XML. It doesn't because the opening and closing delimiters < and > are escaped on GENERATE.
