3.org/TR/REC-xml/#sec-cdata-sect everything within a CDATA block is unparsed except for ]]>
Now we have following problem, that CDATA inserts can happen everywhere not only within the CDATA instructions.
What we have to do now is to double buffer CDATA blocks until their end and also!!! parse their content for CDATA embedding and replace it with an escaped end sequence.
Now parsing CDATA embedding is a little bit problematic in case of PPR because it can happen that someone simply adds a CDATA in a javascript string or somewhere else. Because he/she is not aware that we wrap the entire content into CDATA. Simply encoding and decoding of the CDATA is similarly problematic because the browser then chokes on embedded // sections
What we do for now is to simply remove // and replace all other pending cdatas with their cdata escapes ]]> becomes <![CDATA[]]]]><![CDATA[>
If this causes problems in corner cases we also can add a second encoding step in case of the cdata Javascript comment removal is not enough to cover all corner cases.
For now I will only implement this in the impl, due to the spec stating that implementations are responsible of the correct CDATA handling!
@author Werner Punz (latest modification by $Author$)
@version $Revision$ $Date$