That would only make sense if everyone used a totally consistent amount of bandwidth 24/7. The reality is that most bandwidth is used between 5-10pm. Bandwidth caps cause people to use bandwidth sparingly and ease congestion. To allay your complaints about unfairness, you could structure it so that the caps are looser in off hours like 3am, but then again, that wouldn’t benefit most people because they don’t use the internet at 3am.

The best system would be to charge based on bandwidth usage, and raise and lower prices based on congestion per tower. But most ISPs don’t have the billing capability to support that, and bandwidth caps sort of approximate it.

>To allay your complaints about unfairness, you could structure it so that the caps are looser in off hours like 3am, but then again, that wouldn’t benefit most people because they don’t use the internet at 3am.

On the other hand, something like this might incentivize developers to take advantage of lower pricing during periods of lower congestion. Mobile OSes tend to provide the user a choice between downloading updates anytime or only when on WiFi, but if the pricing structure made this useful, there’s no reason they couldn’t download updates overnight too.

Back in the early 2000s, before most plans move to "unlimited", some ISPs in Portugal did exactly that: they had rather small caps, but then unlimited between 1am and 7am.

And we did have developers take advantage of that; for example, there was a popular fork of eMule that had extra scheduling features, so it could automatically run just in that period.

>I try very hard to be civil on HN, but everyone defending artifical data caps are a bunch of idiots

Not that hard I guess.

Data caps are the only reason you have affordable consumer internet because they allow a significant amount of oversubscription which matches the mostly-idle bursty behavior of consumers.

You can get leased lines from ISPs with no caps easily. You just won't like the real price that comes with guaranteeing a customer that kind of bandwidth.

And why should content creators also be the ones who control the very last mile of the internet in the USA? You don't think there's something very wrong with that arrangement?

Back in the day, we used to take a dim look at companies who amassed horizontal or vertical monopolies, and broke them up. Well, they're like the replicators from Stargate SG1 and have reassembled into something even uglier.

And yet when we discuss this, you fall back on caps and justification by contention ratio... But why are their services zero rated, eh? Of course we know why - it's to kill the opposing services like Netflix, Hulu, or others... except for their services!

Logically, the actual lines should belong to the people (government). The money We've(royal) paid has been above and beyond what it would have cost for a REMC-style internet-ification. I don't want the US or State govts being ISPs themselves, but muni style arrangements have worked out well in the past - except when Comcast and ATT haven't lobbied them out of existence.

> And yet when we discuss this, you fall back on caps and justification by contention ratio... But why are their services zero rated, eh? Of course we know why - it's to kill the opposing services like Netflix, Hulu, or others... except for their services!

I mean there is also the fact that TV and broadband run on entirey separate infrastructure that just happens to share a physical wire. On FiOS, for example, TV is a separate wavelength of laser and is a broadcast signal, not an IP signal.

How can it be entirely separate infrastructure if it uses the same physical wire? Using those wavelengths for that leaves less carrying capacity for internet data. It's in direct contention for the same resource.

The video signal is muxed onto the fiber after the OLT, so the only shared portion of the infrastructure is the passive fiber. That’s effectively an uncontended resource. The single mode fiber has vastly more capacity than you can actually use with the active components of the network. Your limit in data speeds is from the OLT, and the Ethernet ports those OLTs are plugged into, none of which is shared with TV.

"The fiber has plenty of capacity" seems like an argument against data caps, considering that getting/keeping the fiber in the ground is the major cost.

The whole concept of separation seems like a dodge. Is there any real advantage to using separate equipment for TV vs. using the same hardware resources to provide more data capacity and using IP multicast? Or is that what they are doing, and just doing it on separate hardware in order to claim separation?

> "The fiber has plenty of capacity" seems like an argument against data caps, considering that getting/keeping the fiber in the ground is the major cost.

No, a huge part of the cost is the active equipment, which also requires maintenance and upgrades. Have you priced aggregation routers with 10G ports recently?

> Is there any real advantage to using separate equipment for TV vs. using the same hardware resources to provide more data capacity and using IP multicast?

Yes, injecting the TV signal as RF over glass is vastly simpler, cheaper, and more reliable than using IP multicast. There is also the fact that it pre-dates streaming services, and is an outgrowth of the analog CATV distribution systems that were in places for decades before streaming Internet TV.

> No, a huge part of the cost is the active equipment, which also requires maintenance and upgrades. Have you priced aggregation routers with 10G ports recently?

At something like $1000/port, for 100 Mbps symmetric broadband customers that's $10/customer divided by the oversubscription ratio, per hardware refresh cycle (say three years). The percentage of the customer's monthly bill rounds to zero.

And the maintenance cost may be significant, but it also shouldn't really be proportional to the traffic level.

> Yes, injecting the TV signal as RF over glass is vastly simpler, cheaper, and more reliable than using IP multicast.

It just seems like the opposite of how everything else is going. Virtualization, SDN, etc. Get everything running on the same kind of hardware so everything is fungible, failed hardware can be swapped out with universal replacements, idle resources can be reallocated to other workloads without changing hardware, etc.

> There is also the fact that it pre-dates streaming services, and is an outgrowth of the analog CATV distribution systems that were in places for decades before streaming Internet TV.

This would be easier to believe from Comcast than Verizon.

> At something like $1000/port, for 100 Mbps symmetric broadband customers that's $10/customer divided by the oversubscription ratio, per hardware refresh cycle (say three years). The percentage of the customer's monthly bill rounds to zero.

If you're only building enough active equipment to support 100 mbps service, then your point about sharing capacity on the fiber itself with RF TV is entirely irrelevant--you've got nowhere near enough active equipment to exceed the capacity of even a single pair of wavelengths.

Compare that to something like Comcast's Gigabit Pro. Each user gets their own port on a 10G aggregation router. The CPE is a Juniper ACX-2100. And you need a beefy router after that to do traffic shaping. (Due to the way the traffic policing works, Comcast will happily deliver 25MB at 10 gbps before the policer kicks in, which completely overwhelms any consumer router you put after the Juniper.) That's thousands of dollars per user in active equipment--and you're still using just a single pair of wavelengths.

> It just seems like the opposite of how everything else is going. Virtualization, SDN, etc. Get everything running on the same kind of hardware so everything is fungible, failed hardware can be swapped out with universal replacements, idle resources can be reallocated to other workloads without changing hardware, etc.

There is nothing cheap, simple, or reliable about SDN. Flexible yes, but not those other things. It takes enormous amounts of hardware to, e.g. build an SDN switch with VPP that can compare to a $1,000 Juniper 10G switch.

RF over Glass is super easy and cheap. You get a feed from a head-end, and you inject it into a passive fiber network. There is no IP, no routing, nothing active in the delivery network until you get to the CPE.

They don't have to. It's only because the alternative is to invest in CDN, which some don't want to do. Real Internet video providers do that however (Youtube, Netflix and so on).

Either way, it doesn't mean they aren't competing. Legacy TV business hates Internet video, even though they know well, they are finished and either they'll go bust or they need to switch to Internet video themselves.

data caps are not the reason. $5bi of free government money is. Remember that one?

The natural data cap is the line speed of the ISP's peering links to upstream providers divided by the number of customers.

Traffic out to the internet involves way more than your last mile connection.

> The natural data cap is the line speed of the ISP's peering links to upstream providers divided by the number of customers.

The natural data cap is the line speed of the ISP's peering links to upstream providers divided by the number of active customers. But you don't need monthly limits for that, it's true all the time inherently.

It's also kind of a scam on the ISP side because content providers are generally happy to peer with ISPs at no cost to the ISP, so it's not as if there's a non-artificial bottleneck there.

> purely because the internet companies are also content companies

I have no doubt this is a reason, but most ISPs also oversell their bandwidth. If every customer with a 10Mbit line is only using it 10% of the time, your upstream lines only 'need' to be 10% of 10Mbit * the number of customers (plus whatever margin for spikes).

A 100mbit connection that is not oversold can easily run a few hundred to a thousand dollars a month.

Maybe a decade ago. It's down below $50/month now from a major exchange point. Even less if you deal in larger quantities as ISPs would.

It's called contention. Typical ratios are about 20:1 - for every 20Mbits of throughput they sell to customers, they have about 1Mbit of throughput to the internet. Leased lines are uncontended, but they're also vastly more expensive than conventional broadband services.

I don't care how they do contention. That's a business plan.

What I do care about is fraudulent business practices. I expect a minimum speed alongside a maximum speed. And if their contention ratio is 10:1 then I expect 1Mbit-10Mbit for that connection.

But no, the content/internet media companies play insane games, zero rating their stuff, enforcing arbitrary 'kill netflix' limits, and evil layer 7 filtering. They need broken up into lines owned by the state, service over lines sold to whomever provides service (like an ISP or a content company), and customers leasing the lines like how power works.

These megagiant media corps should have never owned the physical connections. At all.

The problem is there's simply not enough capacity for _everyone_ to use their 10 Mbit connection at the same time. That's a necessary property of efficient packet-switched networks.

As a result of this, people who use more total data generally incur higher infrastructure costs to the provider than those who use less; even if those two groups of people have the exact same link speed.