The PCRE has some interesting suggestions on how the grammar for a 
recursive regular expressions might work.  I am concerned about the use of 
(?P>name) to call a regexp subexpression as an atomic subroutine.  The (?
P>name) format has never before been supported by Python and the (?P...) 
notation is exclusive to python, so it is strange the PCRE assigning us a 
use for (?P>name) without the Python community actually agreeing to it.  
Other than that, though, I think this is a possible feature we could add 
in issue 2636 as item 35.