[rfc-i] New proposal for "canonical and others"

[John Levine]

>Assuming that there are multiple formats (this seems unfortunate but 
>inevitable,) then even if they are generated by tooling there will be 
>times when they differ.  Based on experience in other contexts when that 
>happens, it is very helpful if there is a well define "right answer" 
>even before folks get around to fixing the problem.

I've been assuming one form will be canonical/archival/official, and
the other forms will be created mechanically.  If there's a problem
with the derivative version, we can fix the translator and redo them.

R's,
John