How about improving public internet infrastructure instead?
If you don't want solutions like this, are you also against private networks? In addition to Riot, Google famously cuts the gordian knot of CAP in Spanner by declaring their network won't go down.
If it's ok to do it yourself but not outsource it, "think of the little guy" holds no water.
Although it should be noted that all of this effort is only worth it because greedy ISPs are unwilling to peer to certain other networks in most (all?) cases, as the private network is still running through the same cables for most of the way.
That's the logic of the above post. The answer is NEITHER "outsource it" NOR "do it yourself". Instead, if the public offering is good enough, you dont need either option. That's how the internet has worked and grown so far.
Not sure I can get on board with that, but it would level the playing field.
Many of the complaints in the reaction to the article is that the article is using the language of net neutrality to argue for the opposite (pay for access, which includes the ability to pay more for privileged access).
Having a protocol and routes that guarantee low ping but require low bandwidth makes sense to me. A lot of the current infrastructure is built for TCP: something that can handle losses and actually uses packet loss to maximize its bandwidth use.
Having specific shortcuts to open fast lines do not threaten net neutrality. For instance, I think game devs, would love if you could "reserve" 10KB/s of the lowest ping you have, out of your 500 MB/s "bandwidth budget" and reserve it to a given port.
And note that it is not just games that require it. Telepresence applications are limited also because of this problem.
Paying for higher-priority traffic/more bandwidth looks very, very similar to paying for traffic not be throttled. Some might argue it's the same thing.
IMO the difference is that backbone packet transit is different than end-user ISP transit. Unfortunately net-neutrality gets very muddy with, eg Verizon being both the backbone and the end-user ISP. Netflix network engineers are keenly aware of transit, which is part of the reason why they run their own CDN.
This is an optimisation which can be made by the game developer / host, where the alternative would be prioritisation by the ISP, which would go against net neutrality.
I actually think the problem this is meant to solve is an example of why, from a purely pragmatic sense, net neutrality is effectively impossible to deliver at scale. Someone has to eat the costs somewhere, and the publishers are the ones who benefit the most financially.
That's exactly what paid prioritization is though? A less congested bandwidth link free of peasants?
CDNs can get closer to end-users by peering at Internet Exchange Points (IXPs).
* https://en.wikipedia.org/wiki/Internet_exchange_point
Maybe not as close/fast as right on an ISP's network, but probably closer than a trans-continent or trans-ocean hop.
I do not understand this part: why is a business agreement necessary to simply discover BGP paths?
There is an ISP with users that want to go to Youtube or Akamai: the ISP has a router with the full BGP prefix list of the Internet, including for those of YT and A. YT and A are probably in most of the larger IXPs, and presumably the ISP in question is also at a few IXPs.
Why would a business agreement be necessary for the ISP to send traffic over the IXP's switches to the CDN(s)? What's the point of connecting into an IXP if you add the 'overhead' of business agreements?
Because that's how distribution works and has always worked under corporate capitalism: the company making money off the content itself pays to distribute it (on the Internet, this is generally understood to be the originator of the packets). So the ISPs charge the CDNs and companies like Netflix to accept their traffic. A common argument is "well I already pay for that as a customer!" Telecom business models are a lot more complicated than that, and if 100% of their revenue came from subscription fees you'd be paying a lot more than you do now.
> What's the point of connecting into an IXP if you add the 'overhead' of business agreements?
Because it's a convenient location to house network gear with easy access to multiple large ISPs and upstream networks? An IXP is just a datacenter. What makes it an IXP is the fact that multiple large telco networks are hosted there.
From the wikipedia article linked above: "The Vancouver Transit Exchange, for example, is described as a 'shopping mall' of service providers at one central location, making it easy to switch providers, 'as simple as getting a VLAN to a new provider'. The VTE is run by BCNET, a public entity."
But if I am with an ISP A, and I want to watch something on YouTube, then given that I am paying my ISP for connecting me to "the Internet", and YT is on "the Internet", how/why would YT pay the ISP anything?
How exactly would an ISP charge a CDN, YT, or Netflix? The content distributors simply connect to the Internet and advertise via BGP: besides paying their own ISP(s), how would a content provider pay a 'distant ISP'? If a content provider is willing to pay for dark fibre and install their own gear into IXPs, how would any ISP issue an invoice to the CDN(s)?
And the "transit exchange" at VanIX seems to be separate from the open peering option available by simply advertising to the router servers. From the sentence right before the one you quote:
> When these conditions are met, and a contractual structure exists to create a market to purchase network services, the IXP is sometimes called a "transit exchange".
* https://en.wikipedia.org/wiki/Internet_exchange_point#Traffi...
TorIX has the majority of participants doing simple peering:
> The Exchange also offers two BGP Route-Servers, which allow peers to exchange prefixes with each other while minimizing the number of direct BGP peering sessions configured on their routers.[3] Participation is voluntary, with approximately 85 percent of the membership using the free service.
The behaviour of discriminating based on application or financial potential is not tolerated (legally or socially) in other common carriers like mail.
It almost always is a cost savings -- I've worked enough interconnect agreements and the dispute is always over value capture (who gets how much). Transit is expensive; but transit is paid by the originator. The ISPs incur indirect costs that are harder to measure, so it has to be negotiated.
> The behaviour of discriminating based on application or financial potential is not tolerated (legally or socially) in other common carriers like mail.
It's not? Because big companies absolutely get preferential treatment with mail too. USPS does things for Amazon they do not for anyone else.
It hasn't always been like this, but deregulation craze in the 90s/2000s weakened a lot of the protections we had against it.
This is exactly what peering is a significant chunk of the time. You pay for better peering in the fees that you pay to the right server parks. If that's the exact opposite of net neutrality, it has been dead for a very long time.
Transport however is indeed traditionally paid.
As for paying someone to send data destined to their network, that's weird. Unlike transport, they want that data, why would you pay to deliver it to them?
As an example, AMS-IX is completely non-profit, see: https://www.ams-ix.net/ams/about-ams-ix
[Edit: realised I didn't actually answer your question: it is much cheaper, at scale]
If an online service can determine where the majority of its users are network-wise, then they'd set up a mirror at the closest IXP.
We should anticipate some will say no thank you and strike out forging their own waterways.
My customers are delighted by this and in turn they spend more with me and less with my USPS shipping competitors.
Data packets are just digital shipped goods. The same rules should apply.
You get a lot better rate if you can fill a whole intermodal container. You get an even better rate if you buy that containers capacity for an agreed upon period of time.
There's a huge industry built around optimally buying space in shipping containers and transportation.
Or that any unbought priority capacity is made available to neutral internet traffic.
It's not a mutually exclusive situation.
It really is a mutually exclusive situation, either all traffic is treated equally or it isn't.
This isn't animal farm, some packets are not more equal than others
Some packets are more equal because they are being greased with payment to transport providers. Like Bitcoin transactions, you can pay more to have it done faster. Or the HOV lane in big cities - pay a fee or bundle the 'packets' to get downtown ahead of the rest of the traffic.
That sounds like a physical fast lane to me. Pay more, get a faster, more direct route.
Obviously not all packets are equal. VOIP is an example of a high priority protocol. BitTorrent is low priority.
Legislating traffic shaping at this level would be absurd. Have I been living in a cave? Are Net Neutrality advocates arguing that it should be illegal for network operators to perform any kind of traffic shaping, even that which would prioritize the traffic for latency sensitive applications?
At least some of us NN advocates believe it should be illegal for ISPs to perform the kinds to traffic shaping that explicitly identify and prioritize certain ports or protocols over others. Because that kind of traffic shaping is not really necessary to offer good QoS, and thus there's no reason to continue allowing it.
It is nobody on HN's responsibility to educate me on this, but I'd love some good hearty technical reading on this topic, because this is definitely counter-intuitive to me. If anyone has a link to a resource on this topic or feels motivated to type out a technical description of how QoS could work without explicitly identifying and prioritizing certain traffic, I will read it raptly and greatly appreciate the additional education.
CoDel on its own prevents a high-latency queue of packets from building up, but since it operates on a single FIFO queue is indiscriminate about which packets get dropped when it's time to drop something. fq_codel and Cake will tend to give priority to new or sparse traffic flows, and when they need to drop a packet they will drop from a flow that has a standing queue—on the assumption that those high-bandwidth flows are likely to be less latency/drop sensitive and probably can back off on their transmit rate. So any protocol with VOIP-like traffic patterns will tend to get prioritized enough to have minimal added latency and no packet loss (provided it's using a small share of available bandwidth), and the packet drops/ECN markings will hit the network flows that are behaving like TCP bulk file downloads. These heuristics do imbue fq_codel and Cake with a bias toward certain traffic-handling policies, but it's very analogous to the heuristics used by a typical operating system CPU scheduler, and well-grounded theoretically and empirically.
I said at the beginning "for home routers", because these new AQMs have not yet been incorporated into the kinds of ASICs used for carrier-grade equipment. But anywhere that it is practical to deploy these algorithms, they are easier to configure and offer better performance than the now-obsolete QoS strategies that depend on things like trying to decide whether port 53 should go to the head of the line to speed up DNS queries. These new algorithms have proven that ISPs do not need to buy any equipment to do things like detect and throttle bittorrent traffic in order to prevent it from overwhelming their network. They just need to upgrade their routers and gateways to use good general-purpose traffic management techniques.
These tighter margins will (if they weren't government supported) eventually mean that the USPS should reduce the number of trucks they run to better optimize their costs.
No one would have taken on IBM
And if indeed a monopoly were to develop, it’s not as if they could charge whatever they want with impunity. There are many multi-billion dollar companies that would be more than happy to expand into the logistics field if the potential profits are there. FedEx would need to keep their costs reigned in to stave off that threat.
Even if a monopoly could never be assailed (untrue, but for the sake of argument) they still couldn’t do whatever they want. They raise prices, people will search for alternatives. Overnighting a contract too expensive? Companies will start to shift more towards secure verified digital signature and verification methods, and the service providers lose money. Bandwidth being too throttled to play online FPS games or stream HD shows? People will start to look towards more localized or even non-digital options, and the service providers lose money.
Companies like Amazon or Walmart?
The point I was making was a lot simpler: you could fly 1, 100, 100000 additional FedEx priority jets around the world and it wouldn't delay a single USPS truck one second longer.
The original point at the top was that low-latency gaming across the net reduces a QoS issue to a net neutrality issue in the end if we can't add more bandwidth (or airspace) the moment its needed.
And those FedEx planes could slow down the USPS traffic, more FedEx planes means more last mile trucks, more road congestion, FedEx buying more optimally located sorting centers, last mile carriers prioritizing FedEx pickup/drop-off over USPS, etc.
You used the FedEx plane as an analogously to new faster/higher bandwidth infrastructure. The USPS trucks contents may have been on one of those jets at one point (if the contents are considered to be packets, and the truck was a lower bandwidth link).
However, network operators nowadays, instead of building up infrastructure capable of ensuring all traffic gets similar QoS (which costs money), resort to traffic shaping, route analysis, and DPI (looking in the box to figure out what it is), to try to eke out that QoS is more congested links by making the losses at least explain away-able.
This is the opposite of the colloquial way of handling net neutrality, which is to faithfully deliver every packet regardless of content, and mind your business in terms of what it contains.
You don't lookatthe package. You don't shake it. You don't paw at it, poke it, or molest it. Just get it from A to B.
Or did I mess up?
Also I am pretty sure a lot of our infrastructure for the internet was subsidized (via grants or what not) by the government for the big tel-co companies (I have been trying to find legit links to source, but in my quick limited search only found Reddit threads). If this is true, then we, the taxpayers, helped pay for the internet we use and should get at least net neutrality.
Just my opinion, probably wont match everyone else.
Building a shipment company which can reach everyone in a country is a substantial capital investment, which I suspect is higher than that for a nationwide private network, though which is higher is irrelevant to the argument.
The roads, which the trucks require, are subsidized by the government.
In other words, I don't see the differences you're claiming.
Net neutrality, as an argument, applies to Internet; I've never seen a claim that it should apply to private networks, even those which use IP/TCP/UDP.
$50m-ish I think in 70s-80s money before they even shipped a package.
And they still barely made it off the starting line.
Thats more like saying that anyone can pick up a piece of mail and become a mail carrier. There are all manner of legal and practical barriers that require an internet to work outside of your street.
Where I am, it's a simple matter of setting up some high-power relays and building a mesh network in the permitted radio band. People do it all the time.
Also, I think you meant courier. Mail is quite tightly regulated at the federal level, at least in the United States.
FedEx does control which locations they stop at and make pickups/dropoffs.
You can pay more to make sure they only stop with you.
And I believe a friend's law firm has a bulk deal because of how much next day mail they need to send.
People have a right to build private internet infrastructure and sell it at a premium. Full stop. No qualifications.
Your opinion is naive and short-sighted.
You're essentially asking, "why don't you ignore the thing that definitely works right now, and instead try the thing that might work out in a decade or two if you get lucky?" Why would any sane business choose that option?
That's what GPs remark of "only big players can afford private networks" was about, and it's the opposite of what you're claiming.