[selectors-l4] a crazy observation?

I would like to propose something that may seem crazy: please don't lynch
me.

I think we are barking up the wrong tree with the slow selectors  in L4
entirely.  What if we were to remove them altogether and instead make it
possible to create author defined pseudos.  Tab mentioned this on his blog
http://www.xanthir.com/b4N80 (see polyfills section toward end).  Obviously
these would present a kind of single challenge and -may- introduce a new
kind of perf expectation - none of it would necessarily place an
unreasonable blame on an implementer for a specific feature, we could allow
ideas (maybe even limited optimization strategies) to  compete in the wild
and collect data.  A single polyfill (developed and presented with a draft)
could lift all boats in the meantime.