NNSquad - Network Neutrality Squad
[ NNSquad ] Re: EVDO Observation
Bob Frankston wrote: > I don't know if this is relevant to this list or should be a side discussion > but we shouldn't assume that "bits per second" is the only measure. Within > my home I "subscribe" to a copper wire and can run it as fast as I can. > > Are we just measuring service provider promises or can we measure against > available facilities and view arbitrary caps on the use of the facilities > akin to creating measures like "minutes" which are purely accounting > fictions over IP? I think you're on to something here, though you still seem interested in a much broader measure of performance than necessary or relevant to a neutrality discussion, one that could easily take us down a rat hole. This isn't about guaranteeing performance for any particular application. It's not even about whether prices match costs (e.g., your charging for fictional "minutes"). It's about providing the same performance (whatever it is) at the same cost (whatever it is) when certain things change, like packet contents. Think of it as taking a partial derivative. d-latency/d-contents, for example. Your arbitrary transfer cap would show up as a volume-dependent increase of latency to infinity (I know, it's a little strained, but I kinda like my approach of turning everything possible into latency). But you bring up a very important point: how do we distinguish punitive or discriminatory behavior on the part of the ISP from ordinary contention? Since there are rarely any service guarantees, and most ISPs don't publish their provisioning guidelines, it may be hard to even tell when volume-dependent discrimination is occurring. Sometimes they'll openly disclose it, such as a speed cap or monthly transfer cap, but it might also be undisclosed and very subtle. You might have to carefully compare the experiences of several nearby users. It should, however, be fairly easy to devise a good test for content-dependent discrimination. Same for spoofed traffic. I think one of the beneficial things that might come out of a constructive discussion between users, vendors and ISPs is a better understanding of how user-viewed performance changes (or should change) with user load. For example, my QoS proposal involves taking capacity and dividing it by the number of users who have been active in some recent interval (e.g., 30 sec). Each active user is then guaranteed that fraction of total capacity. When someone doesn't use all of his guarantee, it goes back into a pool where everybody else can fight over it. If an ISP were to implement and openly document such a scheme, and put up a real-time indicator of the number of active users, then everyone would know what kind of service to expect. There still wouldn't be any absolute guarantees, especially for the paths beyond their ISP's control, but at least their ISPs wouldn't be the black boxes that many of them now seem to be.