I think when the time comes to revise 2047, you need to take account of
what real implementations actually do, and try to move in their direction.
so we make the implementation bugs a part of the standard? that's nuts.
I think if an implementation has made a useful improvement to what the
standard demands (and if that 'improvement' is widely followed by other
implementations), then it is reasonable for the next standard to take it
on board.
perhaps. but those 'useful improvements' are a tiny subset of
'what real implementations actually do' that deviate from the standard.
Keith