5 ms·

What lies ahead: Net Neutrality

17 points by sant0sk1 16y ago

SkyMarshal 16y ago
Money Quote: "I see the idea of "absolute" net neutrality going away at some point. Eric Schmidt made an important point at Web 2.0 Summit: There are two concepts of net neutrality. One is you can’t discriminate against any particular company or any particular application. But on the other hand you can discriminate against classes of applications."
loup-vaillant 16y ago
I am very disturbed by the idea that prioritizing classes of application is an okay thing to do. First, it's impossible in the long run: if for instance you slow down everything but the web and e-mail, you'll see more and more application using HTTP as a transport layer. This can quickly degenerate into an arms race. Second, it requires the ISP to effectively wiretap their users. Really, if there's not enough bandwidth, just limit the bandwidth. Users will prioritize their traffic themselves.
- shasta 16y ago
  > First, it's impossible in the long run You assume that the carriers won't track usage of different bandwidth classes and bill / cap them differently. Offering different reliability/latency/throughput guarantees for different classes of data seems like a very reasonable idea to me. If you want to download bulk data using the most expensive transport layer, knock yourself out.
  - loup-vaillant 16y ago
    First, I'm assuming that they can't track usage. Now they can, but if everyone encrypts everything, it will no longer be possible. And filtering or slowing down TLS is not really feasible, because of e-commerce. I'm not certain of this point, but I'm reasonably sure. Second, why the hell should one bit cost me more than another? Or come slower than another? Because it's part of bittorent transaction instead of a video stream? Third, why the freaking hell should it be okay for my ISP to spy on me? (Just in case it's not clear: spying is requiered for prioritization)
    - shasta 16y ago
      Ideally, packet class identification would be at the network or transport layer. I don't like the idea of an ISP doing payload inspection ("spying"), either. But your argument about all bits being the same is naive. Do you think it's fair that you should have to pay more for overnight mail or certified mail?
      - loup-vaillant 16y ago
        I think some of your assumptions are technically wrong. The only role of an ISP is to let you send (& receive) IP datagrams to (& from) any machine with a public IP. Now, one packet could cost more than another, depending on its destination, its provenance, and its size. But not its content. Not a chance. For instance, the TCP port number of an ISP shouldn't influence the cost of an IP datagram: they're not even part of the Internet Protocol. Now, if I want to send packet faster than another, I would just use a private network that is not the internet. And if I want to send "certified" packet, I'd just use an SSH tunnel, which will naturally consume more bandwidth anyway. But if my internet provider do those choices for me, I tend to complain.
- pasbesoin 16y ago
  > Really, if there's not enough bandwidth, just limit the bandwidth. Users will prioritize their traffic themselves. This, I think. The rest is smoke and mirrors. And there may be any number of privately developed solutions. E.g. if streaming's a problem in the evening, time-shift some of it to workday/schoolday hours, when that "last mile" to the house is relatively empty. For example, Netflix could have a look at your queue and send the next item or two along in anticipation. It fits my idea, and the more sensible arguments I've heard, for effective regulation: Paint broadly and without discrimination, and let private enterprise sort out the details within those lines. The ISP's have become quasi utilities: Many people (in the U.S.) have at most one or two providers to choose from. Therefore, regulation needs to make the playing field for services level within a provider's domain. Similarly, their prices need to be regulated, as locally and regionally at least they are monopolists, or members of a very limited oligarchy. Most people can't just pick up and move in order to shop for another provider. I don't know about overall costs, but I recall reading a year or three ago that trunk data costs were halving yearly. As the end user, where is the resulting reduction in cost for my use going? (I'm not seeing it.) To infrastructure, or to excessive enrichment of those who are squeezing me for everything they can get? If the providers don't like this, well, maybe they shouldn't have spent so much time and effort squeezing out all competition. (Which they've done and continue to do. As one example, just look at their reaction to any attempted municipal provider plan. Then there's the way ILECs fight co-location of equipment and enforcement of e.g. line maintenance SLA's for competing DSL providers. (Fought? I'm not sure how much life this market segment has anymore, anyway.)) EDIT: Well, I see now that Tim's post is addressing wireless connectivity, whereas I mostly addressed experiences in the lined world. I'm suspicious of arguments that the two differ. There were the same howls about capacity, some years ago, with regard to hardline service. And, SBC -- now ATT -- dragged their feet in a quite disgraceful and disgusting fashion in the Chicago area (at least) until the cable companies started to eat their lunch, including voice service. Then, suddenly, they were able to make something happen on the broadband front. Radios in wireless devices are still -- and increasingly, given technological advances, needlessly -- crippled. Data delivery very hierarchical and without much caching. Prioritization is just the "easy way out" for the carriers. Tell them to stuff it, and make them work harder to produce some real solutions. If it costs them some money, well, I'd argue that's one thing that profits are for. They've got enough to spare.
nitrogen 16y ago
In my ideal world ISPs should honor customer-provided QoS flags, while limiting total bandwidth as necessary. They can also limit high priority traffic to a certain percentage of a user's total available bandwidth (i.e. 50mbit/s connection, only 8mbit/s may be high priority when the network is heavily loaded). Anything beyond that is abusive, and will reduce the amount of bandwidth-intensive innovation coming from startups that can't afford to pay off the ISPs.
TrevorJ 16y ago
The real complicating factor is wireless, since from an infrastructure standpoint it isn't actually the same network - cell companies have to invest in their own infrastructure to provide internet access wirelessly.

17 points by sant0sk1 16y ago

Money Quote: "I see the idea of "absolute" net neutrality going away at some point. Eric Schmidt made an important point at Web 2.0 Summit: There are two concepts of net neutrality. One is you can’t discriminate against any particular company or any particular application. But on the other hand you can discriminate against classes of applications."

loup-vaillant 16y ago

I am very disturbed by the idea that prioritizing classes of application is an okay thing to do. First, it's impossible in the long run: if for instance you slow down everything but the web and e-mail, you'll see more and more application using HTTP as a transport layer. This can quickly degenerate into an arms race. Second, it requires the ISP to effectively wiretap their users. Really, if there's not enough bandwidth, just limit the bandwidth. Users will prioritize their traffic themselves.

shasta 16y ago
> First, it's impossible in the long run You assume that the carriers won't track usage of different bandwidth classes and bill / cap them differently. Offering different reliability/latency/throughput guarantees for different classes of data seems like a very reasonable idea to me. If you want to download bulk data using the most expensive transport layer, knock yourself out.
- loup-vaillant 16y ago
  First, I'm assuming that they can't track usage. Now they can, but if everyone encrypts everything, it will no longer be possible. And filtering or slowing down TLS is not really feasible, because of e-commerce. I'm not certain of this point, but I'm reasonably sure. Second, why the hell should one bit cost me more than another? Or come slower than another? Because it's part of bittorent transaction instead of a video stream? Third, why the freaking hell should it be okay for my ISP to spy on me? (Just in case it's not clear: spying is requiered for prioritization)
  - shasta 16y ago
    Ideally, packet class identification would be at the network or transport layer. I don't like the idea of an ISP doing payload inspection ("spying"), either. But your argument about all bits being the same is naive. Do you think it's fair that you should have to pay more for overnight mail or certified mail?
    - loup-vaillant 16y ago
      I think some of your assumptions are technically wrong. The only role of an ISP is to let you send (& receive) IP datagrams to (& from) any machine with a public IP. Now, one packet could cost more than another, depending on its destination, its provenance, and its size. But not its content. Not a chance. For instance, the TCP port number of an ISP shouldn't influence the cost of an IP datagram: they're not even part of the Internet Protocol. Now, if I want to send packet faster than another, I would just use a private network that is not the internet. And if I want to send "certified" packet, I'd just use an SSH tunnel, which will naturally consume more bandwidth anyway. But if my internet provider do those choices for me, I tend to complain.
pasbesoin 16y ago
> Really, if there's not enough bandwidth, just limit the bandwidth. Users will prioritize their traffic themselves. This, I think. The rest is smoke and mirrors. And there may be any number of privately developed solutions. E.g. if streaming's a problem in the evening, time-shift some of it to workday/schoolday hours, when that "last mile" to the house is relatively empty. For example, Netflix could have a look at your queue and send the next item or two along in anticipation. It fits my idea, and the more sensible arguments I've heard, for effective regulation: Paint broadly and without discrimination, and let private enterprise sort out the details within those lines. The ISP's have become quasi utilities: Many people (in the U.S.) have at most one or two providers to choose from. Therefore, regulation needs to make the playing field for services level within a provider's domain. Similarly, their prices need to be regulated, as locally and regionally at least they are monopolists, or members of a very limited oligarchy. Most people can't just pick up and move in order to shop for another provider. I don't know about overall costs, but I recall reading a year or three ago that trunk data costs were halving yearly. As the end user, where is the resulting reduction in cost for my use going? (I'm not seeing it.) To infrastructure, or to excessive enrichment of those who are squeezing me for everything they can get? If the providers don't like this, well, maybe they shouldn't have spent so much time and effort squeezing out all competition. (Which they've done and continue to do. As one example, just look at their reaction to any attempted municipal provider plan. Then there's the way ILECs fight co-location of equipment and enforcement of e.g. line maintenance SLA's for competing DSL providers. (Fought? I'm not sure how much life this market segment has anymore, anyway.)) EDIT: Well, I see now that Tim's post is addressing wireless connectivity, whereas I mostly addressed experiences in the lined world. I'm suspicious of arguments that the two differ. There were the same howls about capacity, some years ago, with regard to hardline service. And, SBC -- now ATT -- dragged their feet in a quite disgraceful and disgusting fashion in the Chicago area (at least) until the cable companies started to eat their lunch, including voice service. Then, suddenly, they were able to make something happen on the broadband front. Radios in wireless devices are still -- and increasingly, given technological advances, needlessly -- crippled. Data delivery very hierarchical and without much caching. Prioritization is just the "easy way out" for the carriers. Tell them to stuff it, and make them work harder to produce some real solutions. If it costs them some money, well, I'd argue that's one thing that profits are for. They've got enough to spare.

nitrogen 16y ago

In my ideal world ISPs should honor customer-provided QoS flags, while limiting total bandwidth as necessary. They can also limit high priority traffic to a certain percentage of a user's total available bandwidth (i.e. 50mbit/s connection, only 8mbit/s may be high priority when the network is heavily loaded). Anything beyond that is abusive, and will reduce the amount of bandwidth-intensive innovation coming from startups that can't afford to pay off the ISPs.

TrevorJ 16y ago

The real complicating factor is wireless, since from an infrastructure standpoint it isn't actually the same network - cell companies have to invest in their own infrastructure to provide internet access wirelessly.