Hacker News Reader

Comment by PeterStuer | Hacker News Reader

Yet another network provisioner's marketing department trying to spin the 'network neutrality is bad' spiel as if life would be so much better if only those pesky regulators would allow any means for ISP's to extract the maximum profit out of every packet, even if that means creating artificial scarcity.

It's like hearing Nestle arguing how the availability of drinkable municipal tap water is bad for their bottled water business.

View on HN

They're not saying net neutrality is bad, but that it isn't well suited for latency sensitive applications like gaming. Which appears to be true; it's the whole reason Riot has their own network, after all.

The speed of light prevents people on the other side of the planet from having low latency. Networks can’t significantly improve latency at those distances. So, games need to avoid latency being an issue through game design or split things into separate instances.

In the end it’s just a question of where the servers are relative to you and nothing else really matters. Private networks tend to make things worse by having fewer paths available increasing the average minimum distance.

It takes <67 ms for light to travel half the circumferance of earth. From my location in Europe to Chicago it takes 26 ms.

Cables aren't laid out straight, but network infrastructure adds a majority of the cross continental latency.

But gaming from the moon would not at all be feasible using light. I wonder how we could solve this for future space communications.

Short answer is: we can't :/ the speed of causality is not something we can "solve".

When a server sends some state to a player, they player can’t get a response back without a round trip. The same is true in reverse. Further, the path through fiber optical cables is not the strait nor are fiber optic cables a vacuum. Using lasers through a vacuum tube strait from the user to a server on the other side of the planet might hit 134ms of light lag which is already a significant issue for many games.

PS: LEO satellite constellations like Starlink have a lower theoretical latency around the world than fiber as the signals are traveling in space, but a longer path. Significantly beating that is going to require magic aka a useful fundamental change in the laws of physics as we know them.

"In the end it’s just a question of where the servers are relative to you and nothing else really matters."

Assuming a 'fair' network. In absence of Net Neutrality it was "Dear company X, you better pay us or your packets will be delayed/dropped and your customers unhappy", and " Dear customer Y, you better opt for extra service A,B,C,D, ... if you want to Game/VoiP/VoD/Chat ..." and very much worse.

"Private networks tend to make things worse by having fewer paths available increasing the average minimum distance."

This is not how internetwork routing works. Obviously a straight line darkfibre line from A to B will beat any route a generic Internet can come up with. It is not a case of Riot Direct xor 'the rest of the Internet', but selectively supplementing by deploying handover points at strategic peering locations in the internetworking fabric where things can have a large impact.

As you say, Riot Direct is depending on the internet. Sure, adding routes can make slight improvements, but outside of high frequency traders or data centers users are semi randomly geographically separated. This puts a hard upper bound on any improvement of reasonable cost that quickly approaches just adding more routes on the internet.

Having said that, I admit it’s not pure snake oil and provides minor benefits to their customers.

Thing is that both are orthogonal concepts. You can optimize a network for latency or for throughput. A network optimized for latency can still be net neutral in concept and intention. Even an under-provisioned network that tries to fairly deal with optimizing for both low volume low latency applications and high volume latency tolerant applications at the same time while filtering out bad actors can still be net neutral.

Riot has their own network because they want a different kind of optimization than the traditional ISPs and Internet backbone providers aim for.

Network Neutrality was born out of the observation that dominant (read quasi monopoly/duopoly) were actively undermining offers from 'over the top' (meaning offered over the 'generic' Internet service rather than packets over a dedicated managed service offering from the ISP, think Netflix vs your ISP's own VoD offering) service providers by selectively shaping the traffic to degrade their offers. At the time, services such as Skype (pre-Microsoft acquisition) were their primary target as they impinged on the juicy profits from both traditional PSTN and ISP owned VoiP offers (ISP and Telcos were/are pretty tight).

The history of telecommunications has been a perpetual battle to reign in rent seeking natural oligopolies. To pretend no active fairness policies are needed or desirable is denying the obvious.

This by no means negates that lively and continual debates on exactly how to translate policy intentions and objectives into a real network fabric aren't necessary, as fairness is an emergent property relying on very many different punctual choices in configuration and deployment, the impacts of which can be detrimental to the objective both by accident or through deliberate obfuscation.

> Even an under-provisioned network that tries to fairly deal with optimizing for both low volume low latency applications and high volume latency tolerant applications at the same time while filtering out bad actors can still be net neutral.

How? We seem to be discussing QoS. As in ranking and prioritizing different services. Something that is fine for your local network, but has been used to punish torrent users by ISPs. Possible solution the wiki-page [0] describes are over-provisioning (as in never-even-needing-QoS) followed by a bunch of protocol stuff that would be problematic once being payed for.

I'm very much in favor of regulations like net neutrality for e.g. situations with no effective competition. But at the same time, I'd also accept specific connections with special properties for services needing them. A remote surgery wanting a highly reliable connection with consistent latency being the classic example. And also needing high bandwidth, in case your model only allowed for low bandwidth ever needing high priority.

I don't see a way to allow this while keeping a hard-line net neutrality over everything else stance.

[0] https://en.wikipedia.org/wiki/Quality_of_service

"a bunch of protocol stuff that would be problematic once being payed for."

And there lies the rub. It is not that NN means ISPs aren't allowed to manage their networks and have to let it all melt down into a cesspool of fully congested non yielding packet spewers (which is a strawman the industry likes to portray).

If network traffic is shaped transparently and non-discriminatory (for competing services types, e.g. buffering all latency robust VoD traffic while letting all eSports gaming packets skip the queue to increase overall Quality of Experience (quality as experienced by the end users rather than pure IP level QoS is not a problem) to provide a fair playing field for service providers (favoring a YouTube packet over a Vimeo packet because the former cut you a deal is not OK) and customers (dropping Anna's Instagram packets because she did not order the 'social' option for a few dollars more, or more likely, Zero rating Instagram packets against a volume cap whereas TikTok traffic is not excused) , there is no NN problem

Intents in regulation are a thing. Disputes on interpretations and compliance are a normal part of the process.

It's not spin at all though. web devs and high bandwidth users think the world revolves around them since they consume the majority of traffic; but latency is a huge issue in multiplayer games and players and providers are absolutely willing to pay a premium for lower latencies. You sometimes see professional e-sports players actually relocating closer to the servers for better ping.

> they consume the majority of traffic

Do you have a source for that? Offhand, I would assume that video streaming would be an overwhelming majority of internet traffic.

Video streaming is probably the majority of last-mile internet traffic. Along the public backbone is a different story as most video streaming services serve content at the edge.

Their entire business sounds like a way to circumvent network neutrality, by claiming their network of interconnected private networks is something other than the internet.

Ok, and? That’s not a sin or anything.

For what it’s worth there is definitely a market for this stuff. Starlink, hello?

There's definitely a market for human chattel slavery. Some have argued that's not a sin, too.

Just to make sure I parse your words correctly, you are equating private backbones with human slavery?

Yeah that seems very reasonable and obviously not filled to the brim with a ridiculous level of hyperbole.

>Just to make sure I parse your words correctly, you are equating private backbones with human slavery?

No, you are parsing his words incorrectly.

Bullshit.

He is clearly equating the market demand for current-day private backbones to society's prior demand for human slavery. That's a ludicrous comparison.

If you think these two things are anywhere nearly the same, you are a mega dumbshit.

No, he is pointing out that 'there is a market for it therefore it is not a sin' is a shitty argument because there is a a market for blatantly amoral things. You don't understand how analogies work.

It doesn't really seem to be circumventing NN any more than a VPN.

I see no reason people shouldn't allowed to build interconnects not connected to the wider internet and charge for their use.

I'm not necessarily against the concept in every case. But where do you draw the line between that, and just an alternative internet?

Well, what they actually seem to be saying is that network neutrality should protect against discrimination based on user, application, content or other details that are irrelevant to network transit. The argument is that performance requirements are something a bit different and there is a need for market to allow for competition on performance without allowing discrimination based on the other aspects of the traffic.

The idea being that if your ISP owns a game company that makes a popular FPS, they shouldn't be able to charge higher rates for packets that need high performance based on the fact that those packets are carrying content for a competing FPS.

I'm not sure I necessarily find this inimical to net neutrality or a problem. Isn't this more like a way to provision on-the-fly peering agreements based on pricing and latency?

Of course, if as a consumer you have a lousy ISP and the problem is occuring in last few miles, I don't see that this can help.

Why do people confuse QOS with network neutrality? Those are two completely separate concerns. Still baffles me to this day.

The fear has always been extortion: pay or we'll traffic-shape your application into uselessness.

Yeah but my videochat needs low ping, my torrents and netflix - not so much.

I am totally fine with having two lanes - fast for sensitive traffic and slow. Pay as much you consume. This can be mandated to be neutral.

Who gets to control the "sensitive" label and the pricing? How does pricing work in markets where there is only one viable choice of consumer ISP?

One could easily imagine lanes controlled by the end user.

For example, "You pay $60 per month for a 50Mbit cable connection consisting of 1 Gigabyte of low latency traffic and unlimited regular traffic". Then the user can set certain programs to be "low latency" using an app available from the cable company. If they want more low-latency traffic, they can pay extra.

Actual implementation through CGNAT could simply involve having different source port ranges for low latency traffic.

> An app available from the cable company That will be filled with spyware and antipatterns.

Now the key thing is: does traffic not specially marked as "low latency" get artificial latency imposed on it in order to maintain the value of the low latency offering?

You can imagine all sorts of things that would be great, but you have to acknowledge the reality that market forces don't work brilliantly on ISPs in many places.

I feel like this problem only exists in the minds of programmers since in the realm of law you get to take intent into consideration.

What law, though?

That implies a consumer-friendly regulatory regime that can be persuaded to keep up with all the different attempts to game the system. The idea behind net neutrality would be only having to have this fight once to assert that all traffic is of equal value.

Edit: this comment expresses exactly what I want to say already https://news.ycombinator.com/item?id=21762679

Software vendor and end user. Not the ISP You pay 20 euro for 70-100 mbit slow lane and pay couple more for the low latency bandwidth that you need.

isn't that already happened at peering level? Netflix shaping was doing the rounds on the news some years ago.

Netflix is a very special case. When your service comprises 10% of all Internet traffic, a lot of network management things that are normally irrelevant at the application level start to have shaping effects. A typical website won’t be affected at all by, like, Comcast’s peering policies.

Reply on news.ycombinator.com