Hacker News Reader

Comment by kazagistar | Hacker News Reader

Correct. Except this has nothing to do with "specific types of content" and everything to do with "specific large incumbents paying providers to make their competitors slow". If it was just about broad traffic categories, then maybe it wouldn't be as awful of a plan.

Really, ISPs don't want to just be a utility. Imagine if instead of having to charge for service provided, they could have large companies like Netflix and Hulu bid for the exclusive privilege among video providers to not be throttled; the profits would start raining in without having to even innovate on actual product or service provided. How wonderful.

View on HN

Ehhh, not really. I’m pro net neutrality as much as anyone on HN, maybe more so. But recognize that there is a real business point: content delivery cohosted at the ISP is “free” other than electrical cost of the server, which is probably paid for by the content provider, and is a profit center even without kickbacks. Bits routed to other networks cost money and are a cost center. With last mile speeds increasing but backbone data rates not dropping fast enough it can be a real problem. People who torrent or run bitcoin full nodes are probably costing ISPs large amounts of money.

The simple way to solve this is charge larger customers more. Usage caps have been 100% the norm in Australia since the end of the dial up era. Of course we have other unrelated internet issues, (NBN CVC usage cost crippling ISPs, etc) I pay $69.99 AUD for the ability to transfer 1TB of data in total, up and down, over my ADSL2 connection at “full speed” before the ISP will throttle me down to something like 128 or 256 kbps which either way is uselessly slow (insert web page bloat rant here) until the end of the calendar month when the counters reset. If I ran a torrent client uploading lots of bits, all I would do is “burn my allotment faster” thereby creating an incentive for me to not abuse their network, at the same time as me paying a rate proportional to my use of their network capacity.

Usage caps make no sense. In normal countries the actual cost to the ISP of higher usage is negligible. Having capacity and not using it doesn't save any money, and the same equipment that allows for higher instantaneous transfer speeds (the thing they market) also supports more total capacity.

The actual cost to the ISP of a customer using 100% of their connection the whole month is not very large. If only a small minority do that then the amortized cost per customer is negligible. If everyone starts to do that then the ISP is still fine as long as they price plans appropriately -- a 100Mbit plan costs more than a 20Mbit plan.

Bandwidth caps are ridiculous. It's like saying some people drive more than others so to reduce traffic congestion, after you've driven more than 500 miles in a month you can only drive 8MPH for the rest of the month. What kind of sense does that make unless you have some weird agenda like getting people to take a "free" bus which happens to only travel directly to your own store?

In theory they could charge more for higher usage, except that if they charged their actual cost for additional bandwidth then it would be so close to zero as to make little difference, e.g. an extra $5/month for full 24/7 usage. Passing traffic is really not that expensive. The cost of getting and maintaining a wire into your house to begin with absolutely dwarfs it.

You can't simple ignore fixed overhead costs (including capacity to avoid congestion) just because they aren't marginal costs.

> You can't simple ignore fixed overhead costs (including capacity to avoid congestion) just because they aren't marginal costs.

There is no ignoring them, they're just small relative to an ISP's other costs.

It costs millions of dollars to lay and maintain cable. You have to dig up the road. The equipment gets damaged by weather and bad drivers and idiots with backhoes and has to be repaired on a regular basis. You have to staff a call center to provide customer support, pay for all the advertising these companies do, bill customers, process payments, handle delinquent debts, hire a bunch of lawyers, pay all your technical staff to maintain the network, pay rent and utilities, etc. etc.

That's what your monthly bill pays for. Not one of those things costs a penny less if you can convince people to transfer less data. The only thing that changes is that the ISP has to upgrade a switch to one with more or faster ports, which is a fixed cost of tens of dollars per customer per upgrade cycle, and in many cases it happens regardless of usage because the existing equipment has reached end of life or newer equipment uses less power. The added cost is a tiny fraction of the bill. Make it twice as big and it's still negligible.

> People who torrent or run bitcoin full nodes are probably costing ISPs large amounts of money.

As a customer that is not my problem. If I have been sold a 10MB line which has unlimited data then the company doesn't get to complain when I use all of what has been sold to me, the source of the traffic is completely irrelevant.

The issue is that ISPs oversubscribe residential connections massively (say a contention ratio of 50:1) and have done for years, this is now biting them in the ass with high bandwith apps like video streaming becoming the norm.

This doesn't need to be fixed by destroying net neutrality and shaping traffic that they don't get paid for from both ends (the subscriber and content provider), it needs to be fixed by fixing their packages.

But hey, when you can get paid both ways who cares if you're fucking over both parties?

> content delivery cohosted at the ISP is “free” other than electrical cost of the server

Packets delivered to the ISP at an exchange are equally free. Noone is asking for access providers to collect packets in asia and move them to their customer in the US for free. All this is about is that if I pay a transit provider to move my packets from asia to the US, then the access ISP in the US doesn't get to charge me for actually delivering the packet to their customer, or prioritize someone else's packets over mine. There is nothing wrong with ISPs engineering traffic to be delivered to them via the closest exchange possible, they just don't get to abuse their monopoly over their subscriber's line.

> Bits routed to other networks cost money and are a cost center. With last mile speeds increasing but backbone data rates not dropping fast enough it can be a real problem. People who torrent or run bitcoin full nodes are probably costing ISPs large amounts of money.

Which is solved by discriminating between services how?

If outbound traffic is expensive, you have to charge for outbound traffic, not make youtube free (or whatever).

There is nothing wrong with traffic to on-premises hosted servers being free--if everyone can host servers there under the exact same conditions. The problem is an abuse of a monopoly position, not charging based on costs.

> Packets delivered to the ISP at an exchange are equally free. > There is nothing wrong with ISPs engineering traffic to be delivered to them via the closest exchange possible, they just don't get to abuse their monopoly over their subscriber's line.

I don't think the ISP's problem (opportunity) resides in the scenario you're describing. The problem is the top .0001% of Internet destinations congest the ISP generic network interconnects. Hence, these larger vendors (Netflix, Akamai, Cloudflare, Amazon, Google, Apple, etc) either pay the ISP for direct network peering or server hosting to sustain acceptable performance. However, this typically leaves the cost of growing internal network backbone capacity on the ISP. I believe the scenario you are describing would allow ISPs to offset this capital cost to both the big content providers and the biggest consumers.

> Which is solved by discriminating between services how?

The point about torrents and bitcoin is that these applications are particularly burdensome on routers because of the amount of connection requests as compared to a fairly manageable HTTP or RTSP session. As I mentioned above, server co-location and direct network peering is already a thing that many larger ISPs offer on a non-discriminatory basis. However, most new businesses will piggyback on existing cloud hosting (AWS/Google/Azure) or CDN (Akamai/Google/Cloudfront) arrangements.

> I don't think the ISP's problem (opportunity) resides in the scenario you're describing. The problem is the top .0001% of Internet destinations congest the ISP generic network interconnects. Hence, these larger vendors (Netflix, Akamai, Cloudflare, Amazon, Google, Apple, etc) either pay the ISP for direct network peering or server hosting to sustain acceptable performance. However, this typically leaves the cost of growing internal network backbone capacity on the ISP. I believe the scenario you are describing would allow ISPs to offset this capital cost to both the big content providers and the biggest consumers.

I really don't understand what your point is ... how does avoding load on the internal backbone leave which costs of growing the backbone to the ISP?!

> The point about torrents and bitcoin is that these applications are particularly burdensome on routers because of the amount of connection requests as compared to a fairly manageable HTTP or RTSP session.

Uh? How is the number of connection requests relevant to a router?!

> As I mentioned above, server co-location and direct network peering is already a thing that many larger ISPs offer on a non-discriminatory basis.

Non-discriminatory is no enough, it has to be free or at cost. They are a monopoly, and as such non-discriminatory is still a monopolistic price.

> However, most new businesses will piggyback on existing cloud hosting (AWS/Google/Azure) or CDN (Akamai/Google/Cloudfront) arrangements.

Which isn't necessarily that bad, as long as there is competition among CDNs. A CDN ultimately is just a different kind of transit provider.

> how does avoding load on the internal backbone leave which costs of growing the backbone to the ISP?

I didn't say anything about avoiding load. They don't want to raise capital to build their own backbone - so they want to pass off the cost to the largest traffic sources (Netflix etc) and largest traffic destinations (most oactive consumers).

> How is the number of connection requests relevant to a router?!

More connection requests means more CPU usage on a router. More open connections means more RAM usage.

> Non-discriminatory is no enough, it has to be free or at cost. They are a monopoly, and as such non-discriminatory is still a monopolistic price.

Non-discriminatory means they offer all content providers the same prices. Now if you're saying that they can overcharge because there's no competition, then I agree. This is a problem in the short term. But if I understand Pai's logic, this gap in the market will spur competition (and solutions) to emerge through innovation. I'm skeptical about this logic but cautiously optimistic.

I grant you that this is not a satisfying answer when you only have one shitty ISP option like Comcast and there's no hope fore a competitor in the short term. But given the growth of fixed residential wireless and new spectrum like White Spaces it's possible that a Comcast/Verizon would be doing more damage to their own brand (and content offerings) in the long term by mistreating their customers.

Reply on news.ycombinator.com