This is similar to 'Comments split Text into multiple Nodes #546' when considering a complicated document with lots of escaping

<node> some&amp;data ]]> more&amp;data <node>

This could be rewritten (somewhat simplified) but erroneous

<node> <![CDATA[some&data ]]> more&data]]> <node>

and corrected

<node> <![CDATA[some&data ]]]><![CDATA[]> more&data]]> <node>

In any case (the original and final) the user would like to call doc.child("node").child_value() and access the stored data, without complicated logic to iterate over all children and concatenate their values, in particular as this node could be called 'url' and could be given in various different versions, which require currently various different versions of code to access the full URL, coming back I as a user would like to have an options to retrieve all those flavours nice and cleanly de-escaped with a single access point, because at the end essentially only this single value URL is encoded, from a logical/high level stand point, from low level I see the reason why they are split as they are and stored into multiple child nodes.

If this is an acceptable change, I could offer handing in a pull request.