On Fri, 21 Apr 2006, Robin Berjon wrote: > > On Apr 21, 2006, at 00:18, Mark Nottingham wrote: > > Make it a SHOULD and twiddle your CR exit criteria to take it into account; > > in the long term, implementations will sort themselves out. > > My thoughts exactly.
A specification that doesn't accurately represent what implementations do is useless. Make it implementable, take into account feedback from implementers who are telling you the constraints, and don't write a spec that you know will not represent reality. -- Ian Hickson U+1047E )\._.,--....,'``. fL http://ln.hixie.ch/ U+263A /, _.. \ _\ ;`._ ,. Things that are impossible just take longer. `._.-(,_..'--(,_..'`-.;.'