There are issues...
ISPs should not really know what data is streaming into their portion of the network or where it originates. What is being communicated is none of the business of the carrier be they carrying postal mail or transmitting video. In this respect there is a net neutrality issue because in order to charge someone like NetFlix an ISP would have to be prying into private communications.
Ultimately, the users of the network should pay for the network. Both the customer downloading content and the provider (such as NetFlix) streaming the content are 'users'. They should both pay to send or receive whatever they are sending or receiving to the backbone. They are 'peers' in this sense. NetFlix pays their bandwidth providers to push data up to the backbone and their customers pay *their* bandwidth providers to pull the bandwidth down from the backbone.
It makes little sense for the same traffic to be constantly traversing the network. It *should* be cached as a matter of course. However, if NetFlix must encrypt streams for each individual recipient, it will not actually be the same data that is being sent.
If a company like NetFlix is abusing the network by forcing the same enormous volumes across the network over and over, they should be made to bear some financial cost associated with any malfeasance such as refusing to allow caching. Until they bear the cost of some of the inefficiencies they introduce, they will have little incentive to correct their behavior. That being said, we should err on the side of maximizing neutrality with respect to sender/receiver and the nature of the data being moved.
It seems to me that we come to problems like this because bandwidth is limited and much of that limitation is an enforced artificial scarcity to support the old-fashioned revenue models of the communications cartels. Where I live, Bell Canada still charges the unwary as much as $0.91 per minute for a long distance call within the province. [http://www.bell.ca/Home_phone/Long_distance_rates]. Even with our semi-crippled network infrastructure that is better than a 100000 per cent mark up; pretty good if you can get it.
We are still treating EM spectrum, cable and telephone lines as if what they are carrying is pinned to how it is being carried. This has carved up bandwidth inefficiently and resulted in a lack of competition among the different modes of transport. Both result in higher costs for bandwidth.
We need to get everyone on board to create enormous transparent backbone networks that are essentially public assets that are essentially free to use and to remove regulations artificially propping up differences between transport that no longer apply. We also need to have a conversation guided by people we trust. There is much confusion about all this and it is because the waters have been muddied by people who simply don't understand the network attempting to work against disinformation being supplied by ones who do understand it but have a vested interest in the confusion allowing them to stifle competition and charge more for things than they are worth.
The confusion sown by both the genuinely confused and the network cartels means that we can never have a sensible conversation about prioritizing bandwidth. Some things, such as real-time responses to timing signals, keystrokes, etc require low latency. Some things, such as voice communications, require QOS so that there are no interruptions sufficiently long to interrupt communication. Some things require lots of bandwidth, some require very little. Some traffic, such as text messages require very little in terms of quality. EMail does not suffer much if there are longish delays in moving things about or constrained bandwidth. Real-time video conferencing across state lines requires fairly snappy response times and potentially lots of uninterrupted bandwidth. The value of different qualities of bandwidth differ. In order to maximize the economic efficiency of network investment, we need to be able to set different tariffs for bandwidth of differing value. Unfortunately, we cannot trust any of the incumbent network providers not to abuse such a thing and we cannot trust the system overall to protect the disadvantaged from being pushed out into a second class slow lane.
The ideal would be to have nothing but ultra-low latency and essentially unlimited, uninterrupted bandwidth. That is, the ideal would be if there was only one single quality of bandwidth that was adequate for all needs. That is not likely to happen on a real network for the foreseeable future and hence we need to be realistic about how we charge for different types of bandwidth.
There is much that requires improvement on our global network. I don't think that the status quo of ridiculous confusion is ultimately helping anybody. It certainly is not maximizing the greatest good.