Project stance on backwards compatibility

[liamdennehy]

In the list of proposed principles for this project #87 I excluded preserving backwards compatibility, as it is also in conflict with the maxim of "Keep It Simple, Stupid" (also encoded as a principle).

This has wide-reaching impact. We are both fortunate there are already implementations out there as this drives interest and help tease out bugs and corner cases, while at the same time we have a choice to make:

• Preserve all legacy features including those that we have since acknowledged are detrimental, with branching choices and perhaps a version scheme to keep it all tied together (imho way too much bloat) ...
• Aim to publish one, lean specification that is simple to implement and just works, but leaves legacy implementations to fend for themselves.

From my perspective, we are aiming to publish one standard; re-describing and incorporating mis-steps and blind alleys encountered on the way serve no purpose in reaching that goal, and make any eventual version submitted for ratification much too complex.