It seems to me that we are getting things a little out of order here and that is creating a problem.
I believe the situation is quite straight forward. User requirements are established first whether they be for individuals, households, businesses, Government entities, Hospitals, schools etc.
These requirements will vary greatly depending on the services and applications sought under a host of different scenarios depending on location, level of economic development and will tempered by the ability to fund/pay. We can try to determine a set of so called typical scenarios to be used for modelling and then do some straw man technical solutions and costing. If the network architecture is to be based on a shared and thus contended access model, we need to start from average date rates to dimension the network and then depending on the traffic flow dynamics make some predictions about likely peak data rates for individual connections.
There is nothing new in any of this. Thousands of networks have already been deployed in rural areas based on wireless to offer voice and more recently higher speed packet data access. Whether these or future ones are deemed broadband or not is a moot point and not even very important as long as the user needs are satisfied in my view. Furthermore how all this actually relates to NGN in a practical is still not even certain in my mind.
The above view are based on nearly 40 years of practical and theoretical experience in wireless communications, including intimate involvement with wireless broadband network system design and deployment in rural areas.
For consideration by the group.