"requirements documents" considered harmful

> To date, the IETF method of creating requirements documents 
> has often failed badly.

This might be at least in part because

a) IETF groups tend to fail to define the problem they are trying to solve
   in sufficient detail before arguing about "requirements"

b) IETF participants and groups get stuck on the word "requirements" when they
   should instead be thinking in terms of "goals" and "desirable features".  
   "requirements" cannot be compromised; "goals" and "features" can.

I fully agree that we should try to avoid making those mistakes here.