[fluoridation]

I don't think there really is much that can be done. Even under ideal conditions, an ISP could only possibly guarantee the advertised link speed between you and their routers, not between you and any particular node on the Internet. Is it possible an ISP might be doing things that harm the QoS? Yeah, sure. But the angle to approach that problem is not by complaining about instances of limited bandwidth.

[Telaneo]

But the true link speed's not even what's being asked for. 4K Netflix never goes above 20 Mbps as far as I know, so getting just 1/50 the advertised speed to one of the most well-known internet services in existence, hardly seems like a big ask, especially when the only reason that it can't reach that speed or higher is because of the ISP, given that swapping to one that aren't being knobheads about it fixes the problem. It should be the responsibility of the ISP to keep links to other parts of the internet as open as possible. If real-world constraints prevent the speed from being all that high, because it's a shitty server in Australia, then that's understandable. This however, isn't that.

All I'm getting from this is that it's a good idea to label ISPs utilities and bring the hammer down if they're being knobheads about it.

[kbolino]

It is mostly the middle, and not either of the endpoints, that is the real problem. You have a 1Gbps link, the Netflix DC you're reaching probably has multiple links with aggregated bandwidth measured in Tbps, but at some point in between the two there's a 10Gbps link being shared between 5000 subscribers at peak times and now the bottleneck is 2Mbps per subscriber. This link may or may not be under your ISP's (or Netflix's ISP's) control, and it may or may not be the only relevant bottleneck.

The solution that was developed in the Netflix-Comcast fight over a decade ago is content distribution. Instead of trying to build out extra capacity in every possible link, you shorten the path and thus reduce the number of contended links involved in each interaction. This scales much better, but it has two major problems: the first is rightsholders and their obnoxious anti-piracy restrictions, and the second is good old jurisdictional friction and economic misalignment. Somebody has to own the physical servers in all the myriad locations that keep the content closer to the consumer. If the ISP owns them, then they naturally want to exploit them. If Netflix owns them, they naturally don't want to serve their competitors. If a third party owns them, you address those two problems (potentially) but add new ones around liability, non-disclosure, competitiveness, etc.

If regulation is going to be useful here, it needs to focus on opening up opportunities to serve the unsexy middle of the infrastructure puzzle and not just the most visible parts that consumers/voters usually interact with. Also, "Netflix" needs to be understood as just a stand-in for any high-bandwidth Internet service, as the landscape is constantly changing.

[wmf]

at some point in between the two there's a 10Gbps link being shared between 5000 subscribers at peak times and now the bottleneck is 2Mbps per subscriber. This link may or may not be under your ISP's (or Netflix's ISP's) control, and it may or may not be the only relevant bottleneck.

No, that link is absolutely under Vodafone's control. They're deliberately not upgrading it so that they can extort money from Netflix.

The solution ... is content distribution.

CDNs have been worldwide, including Germany, for a long time. That's not the problem here.

[kbolino]

There are two issues here.

If the CDN is so poorly interconnected with Vodafone that there's one bottlenecked link, then it's not really accomplishing its job, at least as far as "inside of Germany" is concerned. It might have reduced pressure on another bottleneck, like links between the US and the EU, but it still needs to spread out more. If Vodafone is blocking that, then pressure should be applied to force them to open up more connections. I'm assuming this CDN serves more than just Netflix, mind you.

Secondly, the question of responsibility cannot be answered the same way today that it was answered in the Internet of universities. Netflix and Vodafone are not peers. The bandwidth ratio between them is incredibly lopsided. This will never change, there is no foreseeable scenario under which Vodafone has a reason to send anywhere near the same amount of data to Netflix as it gets back. This asymmetrical relationship inherently implies a different kind of business arrangement than traditional peering.

What Vodafone (any ISP) provides to Netflix (any content provider) is access to consumers. This is a service, and services are not free. The natural monopoly ISPs enjoy implies some degree of regulatory restraint must be applied on them, but it does not mean they bear all the costs of all the infrastructure either.

However, my bigger point is that this cannot constantly be reduced to these two-party analyses. Netflix is waning, others are rising, this problem needs to be solved in a scalable way.

[wmf]

Nah, you're just an apologist for rent-seeking ISPs and you're trying to cloud the argument with unnecessary details.

[inemesitaffia]

There's no requirement for providers to peer settlement free.

See Chinese providers who will happily buy transit from everyone but make sure it's choked

[kbolino]

"Comments should get more thoughtful and substantive, not less, as a topic gets more divisive."

https://news.ycombinator.com/newsguidelines.html

[wmf]

In the US the FCC named and shamed broadband ISPs for their low speeds and "magically" those speeds increased over the following years. Overcome with greed, some ISPs eventually found ways to cheat on the benchmarks though. https://www.fcc.gov/general/measuring-broadband-america

[fluoridation]

The question was whether a customer could do something.

[wmf]

I suppose in aggregate the customers could use their elected government to fix the problem. In theory.