I really like robert elz's proposal.

As long as the VERSION #define has the right behaviours for mapping the dotted
form into a CPP test on later/earler release forms, I think his proposal has all
the required properties.

I especially like the idea of a version tagging which makes it plain what is
thought to be a more cutting edge yet still 'released' codepoint. Having lived
through the 4.1BSD to 4.2BSD sockets i/f deployment I can relate to the
alternate sub-release often being less stable than you'd wish..

-George