Slashdot Mirror
Europe's 'Net Neutrality' Could Allow Throttling of Torrents and VPNs (torrentfreak.com)
An anonymous reader writes: TorrentFreak reports that the European Parliament is approaching a vote on new telecom regulations that aim to implement net neutrality throughout EU member states. Unfortunately, the legislation hinges on a few key amendments, and experts are warning about the consequences should those amendments fail to pass. "These amendments will ensure that specific types of traffic aren't throttled around the clock, for example. The current language would allow ISPs to throttle BitTorrent traffic permanently if that would optimize overall 'transmission quality.' This is not a far-fetched argument, since torrent traffic can be quite demanding on a network." That's not the only concern: "Besides file-sharing traffic the proposed legislation also allows Internet providers to interfere with encrypted traffic, including VPN connections. Since encrypted traffic can't be classified though deep packet inspection, ISPs may choose to de-prioritize it altogether."
If some ISP starts "de-priotizing" all ecnrypted traffic, they'll soon have 95% de-priotized, which will make it useless anyway.
So they de-prioritize things. That's fine. The competition between ISPs is enough to have some cater to the edge cases. So long as they don't sell a "prioritized" VPN service above what anyone else can provide on their network, I would be happy with the "problems" listed in the summary. They aren't problems, and are fair and equitable.
Learn to love Alaska
Since almost all Google traffic is served via SSL, it will be deprioritized. Sneaky.
If they are allowed to set priorities for different traffic, how is this a net neutrality bill?
How exactly is torrent traffic impactful on an ISP network? they're just routing packets around (okay, maybe you need a larger routing table?), it's the nodes that have to do most of the work. Unless they're using carrier-grade NAT, in which case get IPv6 working you lazy b*s.
Also, looking forward to seeing http encapsulated VPNs!
The torrent part I agree with: torrenting can be very demanding on the networks, and torrents are not used in applications that require real time. They'll be fine if their file transfers take a little bit longer. The encrypted traffic part though - a lot of traffic nowadays is encrypted, so that hardly helps. Furthermore, I don't think that punishing traffic that is encrypted is very fair: the performance overhead is not that great, and I don't understand the obsession with people wanting to monitor and inspect everything. Even in Germany it's a pitiful state of affairs, though not as bad as the US or England. Watching me browse Slashdot is supposed to further secure the state, ja? I don't feel any more secure, and I doubt anyone else does either. I'm very surprised they are as free with it as they are actually; although all (I hope!) banking sites are encrypted nowadays, if they were to read my bank statements unencrypted, I believe that may expose them to a lawsuit from myself...?
"Set a man a fire, he'll be warm for the rest of the night. Set a man afire, he'll be warm for the rest of his life."
Net Neutrality is needed in the US because there's essentially no competition. It's a regulation on a monopoly operator.
Many European countries have competition in the telecoms sector. Any action perceived as unfair throttling will see their customers go elsewhere.
The problem is, regulation is a blunt instrument. If I want decent broadband speed for Netflix, I don't care if everything else is slower. However, it might be in Netflix's interests to offer ISPs a cut to allow higher broadband speeds for its service only. Beneficial to the ISP, to the customer and to Netflix. Strict net neutrality doesn't allow this. Make an exception and you end up with loopholes, and I'm sure there are other potential scenarios where you simply don't want neutrality.
and start deprecating ALL unencrypted protocols.
Establish a new connection dispatch service that all all other services would use. All interconnections would first establish a connection to the dispatch, which would establish a TLS or PGP type of encrypted session, and THEN transit information about which service to connect to.
Whether the governments like it or not, the use of VPN and encryption is on the rise by businesses around the world
My companies, for example, rely on VPN and encryption for all inter-office data traffic, and if EU starts to de-prioritize VPN and/or encrypted traffic many business communication will be hit
Muchas Gracias, Señor Edward Snowden !
Apply priority to 95% of clients and priority doesn't mean anything anymore
Actually it does, but probably not in a good way: it means the other 5% of clients are losing out, perhaps heavily.
Avoiding this scenario -- keeping in mind that a huge proportion of all Internet traffic is generated by a relatively small number of businesses today, and all the little guys between them might only make up 5% of total traffic -- is a large part of why Net Neutrality matters.
If you disagree, post your argument. (-1, Overrated) isn't your personal censorship tool for views you don't like.
I predict more vpn services operating on port 443
It's about time we update bittorrent with better security/encryption and something less easily detectable.
We will all make our own net neutrality when everything is encrypted and nothing can be prioritised. Pros and cos but it's better than "net neutrality" IMO.
A 'singular oddity' is an event that cannot be explained and only happens when you are alone.
The trend is for everything to become encrypted, anyway - so the whole thing will be moot.
Even our company's website defaults to https and we're not even a tech company. YouTube defaults to https. Google. Farcebook, Reddit. (Slashdot seems to be one of the few that don't).
If they start throttling a protocol, people will start making it look like https to work around the throttling - use port 443 and TLS 1.2.
Oolite: Elite-like game. For Mac, Linux and Windows
Really, a bunch of torrenting whiners whine because the proposed regulation acknowledges that ISPs have a legitimate interest in QoS, and that this might impact their download speeds.
Get in your thick heads: Net Neutrality is not about freedom to download, it is about forbidding discriminating traffic based on the endpoints.
"I know I will be modded down for this": where's the option '-1, Asking for it'?
Net neutrality is all about making sure the traffic is not filtered by content, what packet you have on port 80 should not be prioritized because it is coming from cnn.com while the one from say, google.com is throttled because they did not pay an extra fee. It is also about making sure too that the content of the packet is not what decide the throttling, but the functionality and network status. IOW throttling not based on content and origin/destination.
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
This sounds like a bunch of Galileos telling the Pope and his criminal entourage that they need to let people believe the Earth is round. Good luck. I think you're gonna fail because the current profiteers give too much money to politicians, but I admire your tenacity. Verily.
It sucks, but is still better than the present non regulation.
SSH connections as VOIP traffic, Site to Site VPN as Quake 3 multiplayer traffic, https as http but with more overhead and ingenious headers
Routing packets in the order the arrive makes it worse for EVERBODY, and makes very low bandwidth uses like ssh and voip more or less useless.
Streaming video (Netflix) requires a certain (high) BANDWIDTH to avoid repeated buffering. Any more than what it requires does little good, but it needs to transfer X MBs per minute in order to keep up. Latency and jitter do no matter at all for Netflix. It's purely MB per minute- packets can be delayed 200ms and it doesn't matter as long as they arrive before the buffer runs out.
Voip needs very, very little bandwidth- 64Kbps is enough. That's 1% of what video uses. But voip can't have high jitter (variation in latency). It also requires reasonable latency, but jitter is the main issue.
If you have Netflix and voip traffic going through the same router, it doesn't affect the video viewer AT ALL to have a 64 byte voip packet occasionally jump to the front of the queue if it's been waiting too long. Having the voip wait for three seconds of video -would- mean the call goes silent for three seconds. That would be stupid. Really stupid.
Ssh needs virtually zero bandwidth- bytes per second, 1/1,000th as much as video needs. Ssh doesn't care about jitter. But it DOES care very much about latency. When you try type "cat /etc/resolv.conf" it's really annoying to have delays between each character. But the ssh packets are tiny - just a few bytes, so they don't effect anyone else on the network. Again, leaving them waiting in line hurts the ssh user with absolutely no benefit to anyone - it's only damaging. Doing that would again be really dumb.
Suppose a provider has incompetent admins and does ruin ssh, voip, and other low-bandwidth highly interactive traffic by making those packets wait for high-bandwidth non-interactive traffic. People who care about interactive traffic will find that provider's service more or less unusable and switch. So here's a guy (like me) who was using less than 1kbps for ssh while paying the same $45 you pay while you use Netflix. The ISPs cost to service both of us is $70 ($10 for me and $60 for you). Guess what happens when the voip and ssh users leave for a different ISP? We're not there to subsidize your cost anymore, so your bill goes from $45 to $70.
To turn back to your road analogy, you may have noticed that in many places trucks aren't allowed in the left (fast) lane and in most places the left lane is for faster traffic only. If on one tollway all the cars had to line up behind the semis, while another road allowed them to go faster, which road do you think the cars would use? Once the trucks had to pay the full cost of the road by themselves, do you think their toll rates would go down or up? Would the trucks somehow benefit from making it illegal for a car to pass a truck?
Your analysis has one key flaw. It is based on the assumption that there isn't enough bandwidth to keep latency low for everyone. There are natural bottlenecks in an ISP network. A subscriber's ADSL might only be capable of 10Mbps, so as long as the upstream pipe is big enough to handle their constant 10Mbps of streaming video without packet queue depths getting long enough to add more than a few milliseconds to an arriving SSH packet everything is fine.
So while there is an argument for some limited prioritization of traffic, e.g. DNS requests, it is really just duct tape covering up a more fundamental problem. It is also wide open to abuse, because SSH is used for SFTP and a variety of other bandwidth-hogging protocols, and because it is difficult to tell one type of encrypted packet from another. In reality, if an ISP tried to prioritize SSH they would also prioritize VPNs and encrypted P2P traffic. There are some really expensive ways to do it, but why not just spend the money on a bigger pipe instead of a never-ending battle to classify encrypted packets?
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Routing packets in the order the arrive makes it worse for EVERBODY, and makes very low bandwidth uses like ssh and voip more or less useless.
You shouldn't prioritize SSH traffic up, because if you do that, then people's SSH tunnels get prioritized up. You should just have a network without shit latency, which is not massively oversubscribed. You can always prioritize down the stuff you know is streaming. The stuff that goes up is DNS, SIP, and maybe NTP would be nice and the traffic is minimal. And the beginnings of HTTP connections, but only the first however-many-kBs-make-sense.
To turn back to your road analogy, you may have noticed that in many places trucks aren't allowed in the left (fast) lane and in most places the left lane is for faster traffic only. If on one tollway all the cars had to line up behind the semis, while another road allowed them to go faster, which road do you think the cars would use?
No, you cannot use a road analogy like that, because networking doesn't work like that. All the cars on a network link move at the same speed. They're not like automobiles, they're like train cars. Unlike normal trains, the cars are of different lengths, or you could think of them as numbers of contiguous cars with contents belonging to the same customer, and naturally the cars preceding yours will reach the station first. If your cars are going to get there before the ones currently ahead of them, they're going to have to be put on a different line which will get there quicker.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Your analysis has one key flaw. It is based on the assumption that there isn't enough bandwidth to keep latency low for everyone.
That's not an assumption, that's a very real fact.
ISPs design their infrastructure to keep it that way too, and for frankly very sensible reasons.
They should prioritize VPNs.
Or at least offer a work from home package that does.
Added latency in my VPN would have me leaving my ISP.
They can't provide 10mbps per a second all of the time to everybody, well they can, but they wouldn't be able to sell it.
That type of bandwidth is expensive, that isn't to say that ISPs aren't over sold, only that a certain amount of overselling is necissary.
Wow, sent an e-mail as suggested when clicking on "use classic" banner, and got a fast response that addressed my msg
I pay $45/m for a dedicated 100/100 fiber connection to my home here in Midwest USA. My max sustained 1min average bandwidth is 99.9Mb/s. My typical sustained bandwidth when downloading or uploading torrents is 99.5Mb/s, and my min is about 99.25Mb/s. These aren't iPerf tests, but HTTPS/FTP/Torrent. Without doing any traffic shaping, while saturating my connection, my ping never jitters more than 30ms. When I do my own traffic shaping and rate limit to 99.9Mb/s, my ping never jitters more than 1ms.
My max ping to any datacenter in the world is about 250ms and under 5ms of jitter. Upstream provider of my ISP is Level 3 Comm. My ISP guarantees 1:1 internal bandwidth and typically has about 6:1 of trunk bandwidth based on 95th percentile. They are under-subscribed.
I actually did a 120Mb/s external DOS stress test on my connection. The ping to my ISP stayed under 30ms the entire time, but I did have about 20% packetloss. They use a fair-queuing AQM for both up and down to fight bufferbloat.
I've done 1 week long pings to datacenters in LA, London, and Germany. All of them over a 5,000 mile round trip. Less than 5ms of jitter to Europe and under 1ms to LA, and 0.0001% packetloss. My friend was an admin at a datacenter where they had dual DS3 connections to AT&T. He got a 200ms ping to Hawaii and some nasty jitter. I told him I got a 140ms ping and virtually no jitter. He didn't believe me until I got some samples. Even Japan is only 160ms.
Now what were you saying about a "fact"?
Congratulations, you found a good ISP.
If you can prove to me that all ISPs are like yours, I'll concede the point. Until then we both know that my factual statement remains accurate.
If and when big ISPs do disgusting things like blocking access to specific servers or throttling traffic, they deserve to lose your business. Go with the small ISPs run by geeks like us who care able open, free unrestricted access to the Internet for a low fair price.
Repeat after me: http://tgtechnotes.blogspot.co.uk/2014/06/repeat-after-me-net-neutrality-is-not.html
I think there's a mistake here. We have two different applications: * High-bandwidth, latency-insensitive, jitter-insensitive applications, such as torrents, streams, large HTTP actions, etc. * Low-bandwidth, latency-sensitive or jitter-sensitive applications, such as VOIP, SSH, etc. We're trying to jam both of these into the same system and either: a) Fuck everything, no priority because the ISP can't be trusted b) Let the ISP prioritize whatever they please But why? Why not just provide two lines to every user and let them decide? One high-priority low-bandwidth not-oversold line. One low-priority, high-bandwidth, oversold-as-shit line which is thus affected by peak usage, etc. Users can decide what traffic really is high priority and what traffic is garbage. Alternatively, you could provide the same oversold bandwidth to both (ugh), but cap the shit out of the high-priority line to create an incentive not to prioritize shit you don't actually need, fucker. Thoughts?
The two things that are unfortunately used interchangeably, but really are not are Net Neutrality and QoS (Quality of Service). The increasingly multi-service Internet needs QoS to work properly. There are two basic classes of traffic:
1. UDP - This is 'connectionless' in the sense the protocol itself does not handle flow control nor packet recovery. This is generally used for latency sensitive, low to medium bandwidth applications.
2. TCP - This is 'connection' based in the sense the protocol tracks connections at the end points (not the network core), has built in congestion control (tends to use as much bandwidth as it can when asked to), and does lost packet recovery. This is generally used for latency insensitive, often bandwidth intensive applications, though the ACK (acknowledgement) packets are latency sensitive and need prioritization.
With all of the things going on over the Internet now what is really needed goes as follows;
1. A basic 50/50 split between UDP and TCP traffic (or whatever split works best at the time) when the link is congested with the ability to borrow when one side is not using its allocation. The notion here is the UDP traffic is unimpeded unless there is some massive abuse of UDP (which can and does happen in DDoS attacks before they are contained).
2. For the TCP side, TCP ACK packets are prioritized. This keeps heavy traffic going one way from impeding traffic going the other way. You could also set a 50/50 split here with borrowing so DDoS style abuses of this policy are curbed until the DDoS can be stopped.
3. On the UDP side flag all VoIP traffic with a higher priority up to say 30% of the UDP traffic. As usually VoIP uses little bandwidth, but is highly latency sensitive, it makes sense to ensure this traffic can get through.
4. Below VoIP prioritize DNS traffic for up to 10% of the UDP traffic on the pipe. This way your DNS request will be guaranteed to be speedy as long as there is not a DNS reflection attack or the like in progress.
5. Below this prioritize interactive UDP gaming traffic for up to 40% of the UDP traffic. As online gaming does need low latency, but is less critical to our needs than DNS and VoIP, it gets a lower rung.
6. All other UDP traffic can consume what is not used by the above.
7. Do a port check for HTTP/HTTPS and make sure it can get at least 50% of the TCP bandwidth.
8. Do a port check for SSH and make sure it can get at least 10% of TCP bandwidth.
9. Use something like Stochastic Fair Queuing to make sure no TCP streams are really stealing more bandwidth than other streams.
I actually have these rules more or less implemented on my home firewall right now and it has worked out quite well so far. I suppose the biggest difference is I give SSH more interactive capability, but this I consider a personal thing that is much more difficult to resolve a proper implementation on a more universal level. To a large degree ISPs need to keep their backbone lines free of excess congestion for a good user experience, which can be done with current technology, but QoS is also essential as various protocols will step on each other.
Will your Muslim masters even allow you eurotrash servants to use the Internet? Seems like only trouble could come of that.
European men won't need Internet porn once they've been made into eunuchs.
And people are surprised? That's because "Net Neutrality" isn't about net neutrality, whether at home or across the pond. It's a political implementation to take away choice and competition. Stop supporting regulation!
Again, there's no good reasons for non-encrypted traffic. No one should be having to decide just what bits of their online life they are ok with businesses and governments picking over for information to use against them, and which things they are not ok with that. Especially since humans are the worst at recognizing in the short term what will be relevant to those malignant interests. At what point did we decide that governments have the power to decide whether or when people are allowed to have private communications between themselves?
'Just encrypt everything' you said.
They will just de-prioritize anything they can't read and spy on with DPI until it's throughput is useless.
It's time to take a look at some serious steganography.
His ISP is proof that it's possible. If some other ISP is worse, than that is their failure, not a problem inherent to internet routing. Prioritization is only ever needed when the network is congested. Keeping the network virtually free of congestion is the ISP's job. If a network is regularly congested, like many access networks are in the evening, then that is not an indication for the necessity of prioritization, but a sign that the ISP oversells their network and needs to invest in more network capacity! (Just to make this clear: This does not mean that everybody needs to have their bandwidth reserved throughout the entire ISP network. It only means that the contention ratio has to be low enough to statistically avoid regular congestion. Yes, that does mean that you can't oversell bandwidth to a number of clients which is too low to statistically even out the traffic requirements.)
Your analysis has one key flaw. It is based on the assumption that there isn't enough bandwidth to keep latency low for everyone.
So his analysis is flawed because it's based in reality? That's quite a condemning indictment. You are wrong because, um, you are right, but that's not the way I'd like to see it.
If that's the case. No ISP runs a network like you'd like. The obvious solution is for you to build or buy an ISP and run it the way you'd like. The flood of new customers to you would make you a billionaire. Unless you are wrong, and that's why nobdody does it. The 1/10 of 1% that would care don't make enough people to justify the additional cost. A classification machine is cheap. $50k to $100k for one that'll do 40G. But buying 10G more pipes everywhere would cost more than that in interface cost alone, not even counting upstream bandwidth.
Learn to love Alaska
Now what were you saying about a "fact"?
You didn't post any facts. Unsupported statistics aren't facts. Why didn't you name your ISP, if they are so great? Are you one of the 1/100th of 1% that can get Google fiber, or something like that?
Learn to love Alaska
It is also wide open to abuse, because SSH is used for SFTP and a variety of other bandwidth-hogging protocols, and because it is difficult to tell one type of encrypted packet from another.
Which is why the higher priority should only be given to the first x KB of packets each second. Same would benefit VoIP. But yes, it's nearly impossible to distinguish from even HTTPS traffic, since it's encrypted.
If you prioritize the first X KB / sec. worth of packets, you can give a de facto priority to low-bandwidth uses, without giving extra favor to things like tunnels. I don't know if that's being done by anyone. Each additional packet in a given time frame would get a lower and lower weighted priority based on how long it's been since the previous X packets. Might be too CPU intensive for a huge amount of traffic, but probably not worse than deep packet inspection.
So government takeover of ISP's would result in traffic manipulation? NOWAI!
How is that even relevant? Do you doubt that ISPs like that exist? I can assure you that they do exist. For example, three million subscribers in the Netherlands, a country with 17 million inhabitants, have fiber to the home connections and can get symmetric 100Mbps. This network started as a local ISP, and there is no reason why a similar service couldn't exist in the US. That network operator has also begun offering FTTH in Germany and explicitly offers unshared bandwidth throughout the network (I assume they don't actually have enough bandwidth on the backbones to supply full speed to everyone at the same time, but do manage the network to be practically congestion free). They make a point of controlling the actual fibers all the way to the peering points.
Anyway, the point is that a provider can certainly provide a congestion free network under all regularly occurring circumstances, and a congestion free network does not need prioritization. It's as simple as that. The thing which complicates matters is that providers don't want to do their job, and instead prefer to oversell their bandwidth, both on the shared last mile (cable providers) and on the backbones (everybody). They want to keep doing that and throttle "bad" traffic, so that their users don't complain about their phone calls dropping out and their video streams buffering. If they sold 5Mbps instead of 50Mbps, then they could actually deliver what they promise, but of course that would put them at a competitive disadvantage compared to providers which actually deliver high bandwidth without throttling (which is what prioritization really is.)
How is that even relevant?
Because the cryptic nature of the claims, from an undisclosed location in the Midwest, for all we know, the ISP doesn't exist, and he's lying to prove a point. If he gives the ISP, we can look at other 3rd party evaluations for verification.
Yes, I know there are places with good ISPs. The US isn't one of them. And an ISP like Google isn't a reasonable answer, as the last numbers I saw, about 1/100,000 people had access to it, and that was an exaggeration. They need to grow by about 2 orders of magnitude to be considered an option.
Learn to love Alaska
Again, how is that relevant? There only needs to exist one example of a real world ISP to prove that it is possible to run a network congestion free. ISPs like that exist, and they're not even rare. Just because the big US ISPs cheat their customers doesn't mean we have to accept their practices as the best that can be done. Prioritization is a euphemism: It really means someone's packets get throttled and dropped, so that someone else's packets can skip to the front of the queue. That is an unacceptable deviation from fair queuing, and it is only "necessary" because providers allow congestion on their networks, which means they oversold their bandwidth. If the rules were actually fair, these ISPs would be taken to task for not providing the advertised bandwidths to the best of their ability. Taking the money and intentionally not delivering the advertised service is fraud! Congestion on their network is not due to an external force or due to a technical impossibility: It's bean-counters not making the necessary investments. As I said: FRAUD. Why should we accept that?
You stated it as a fact. All I have to do is fine an exception to make your claim false.
Reading comprehension challenge there then. I stated that ISPs do X. As long as more than one ISP does X, you haven't refuted my statement.
Just because 1 doesn't do X is totally fucking irrelevant.
Have a great weekend.
This smells of another government attack on encryption, ALL encryption. It seems governments all over are so intent on surveillance that they will break anything to get it.
And so, what could possibly go wrong with this deprioritization of encrypted traffic?
- No chance of your banking app sensing problems with traffic and either terminating or restarting sessions? I know, there are few reasons to do that, and none technically sound. Assume, for the moment, that your bank has control over how their app works. Now assume you cannot know if 'your'* government has forced them into adding in some interesting quirks. Not outright decryption or backdoors, but perhaps reducing the encryption level in response to "network load". Your ISP is in on this, with a FISA order to deprioritize well-encrypted packets. No matter the source or destination, and certainly no matter the actual network status.
- In the midst of an industry-wide effort to get everyone with a site that uses credentials to go https, this is contrary. But reversing that trend sure gives 'your'* government an opportunity to capture your credentials to all sorts of sites, from the mundane to the actually important (to you). What's the big deal? Well, if 'your'* government would like to keep tabs on your online presence, such as posts to pro-freedom sites, etc, it sure is easier if they can ascertain your identity, and having your login credentials is helpful in that effort. Why would they want to do that? Are you keeping up with US Justice Department instituting the 'Domestic Terrorism Counsel'.
Trust no one, certainly not 'your'* government.
* 'your' government isn't yours any more if it considers you the enemy.
deleting the extra space after periods so i can stay relevant, yeah.
I buy dedicated, guaranteed bandwidth from more than one provider. The cost is around $20/Mbps, at the provider's POP. A line from my office to the POP is quite a bit more expensive.
Since at home I'm only using bandwidth 5% of the time, it would be silly to pay for it 100% of the time. It makes much more sense to share the cost with my neighbors , who also need it only occasionally. Remember on the web you're only using bandwidth while the page is loading, so if my neighbor spends an hour a day surfing the internet, they might be actually loading pages for ten minutes per day. If you're going use the pipe 10 minutes per day, why would you pay for it 24 hours a day? That would be silly. Sharing makes sense, big time. Sharing also means that occasionally multiple people will want to uee it at the same time. I'm fine with that since it's the sharing that makes my home connection $1/Mbps rather than the $20/Mbps I pay for dedicated.
Again, how is that relevant? There only needs to exist one example of a real world ISP to prove that it is possible to run a network congestion free.
The claim that because someone somewhere in the world does it that it's practical in the US regulatory environment is a silly claim. If it were as easy as you claim, why can nobody name any in the US that act in that manner?
Taking the money and intentionally not delivering the advertised service is fraud!
Your Term of Service clearly lay out a "best effort" service, and that's what they deliver. That you disagree with them on the definition of Best Effort indicate you need to sue them (and lose) in court for clarification. Your poor English skills don't trump the law.
Learn to love Alaska
Customer service costs more than bandwidth or infrastructure combined. You may want to revisit your logic about wasting huge amounts of money trying to screw with customers.
No, current DPI platforms have no
Your hypothesis was that "ISPs design their infrastructure to keep it that way [latency low] too, and for frankly very sensible reasons." His ISP is proof that the reasons are not sensible, but selfish. It is possible to achieve low latency without throttling arbitrary protocols.
The advertised service is to provide an "up to" bandwidth, but it has already been established in case law that delivering far less bandwidth is not acceptable, because consumers have a legitimate expectation to get at least x% of the advertised bandwidth (where x% varies from case to case, but is never lower than 50%). I would argue that it is also not "best effort" if the provider only fails to provide the advertised speed because of bean-counting.
We can continue this discussion in German if my English skills offend you.
Having worked at a number of ISPs and you are simply wrong. You have to have pretty bad service to spend anywhere near infrastructure plus bandwidth on your customer service.
And they aren't "trying to screw with customers" they are trying to make the service the best possible for the most number of people. And one of the easy ways to do that is to put in a DPI and set it to put bittorrent as the lowest priority. The people downloading all day long never complain. 99% of the time, they are doing something illegal, the other 1% of the time, they recognize they are using a protocol that's mostly for illegal downloads, so they put up with the slower speeds. Though the best use of a DPI is to run a rolling 30-day count of all users, and put the top 1% of all of them in the naughty bin. That's fairest for all, and works out pretty well. Those using 100% of the pipe 100% of the time and buying a residential service, and complaining on every forum they can find that the cheapest residential service they can find doesn't give listed speeds at all time are better fired. Fire your customers is a business philosophy that pushes your high-cost customers to your competition, and let them deal with them. It works well in ISPs. Save your costs, and have the complaints be about the other guys.
Learn to love Alaska
An ISP must be incredibly wasteful with money if customer service doesn't cost more. The average real world costs of fiber is nearly 1/5th that of copper. Maybe your ISPs were using copper? Even nodes? Nodes in the field are expensive. The cheapest design is a flat model where customers plug into fiber aggregators that plug directly into the local core router.
I've read article after article over the years where they said incumbent ISPs are incredibly wasteful and bandwidth and infrastructure is relatively cheap over the lifetime of a customer. When talking to an ISP network admin, he told me they tried doing QoS and traffic shaping, but issues always cropped up and customer would complain because of poor performance. The cost of handling customers was too expensive, so they just went 100% dedicated. Every customer is guaranteed to have a congestion free, with no caps, traffic shaping, or QoS. Just pure unfettered bandwidth to brute force the issue. Costs went down, customer satisfaction went up.
The only expensive part of correctly designing a network is the up-front cost. But after 3-8 years, that cost is dwarfed by ongoing costs.
Google for Cake and fq_CoDel(cake's ancestor). Stateless AQMs that maximize bandwidth and minimize latency. Virtually no configuration and does 99% of what everyone wants.
Post midnight maintenance. uhggg. https://goo.gl/LlXKAw
tracert www.google.com
Tracing route to www.google.com 216.58.216.100
over a maximum of 30 hops:
1 <1 ms <1 ms <1 ms pfsense.localdomain 192.168.1.1
2 1 ms 1 ms 1 ms 209.xxx
3 7 ms 8 ms 7 ms xxx.Level3.net 4.59.66.x
4 Request timed out.
5 8 ms 8 ms 7 ms Google-level3-50G.Chicago.Level3.net 4.68.71.174
Speeds up to 1 Gigabit. It’s all dedicated symmetrical fiber, so speeds never go down or change.
70 Mbps Dedicated Symmetrical $29.99/mo.
Online gaming
HD streaming
100 Mbps Dedicated Symmetrical $44.99/mo
Web hosting
Heavy online gaming
Online backup
250 Mbps Dedicated Symmetrical $99.99/mo
Cloud computing
File sharing
Heavy online backup
500 Mbps Dedicated Symmetrical $249.99/mo.
eCommerce
Heavy File Sharing
Webinar hosting
1 Gbps Dedicated Symmetrical $299.99/mo
Ultimate eCommerce
Ultimate file sharing
Ultimate cloud computing
Ultimate online gaming
Open Internet Policy
The Company does not unreasonably discriminate in its transmission of lawful traffic over the broadband Internet access services of its customers.
The Company does not block, impair, degrade or delay VoIP applications or services that compete with its voice services and those of its affiliates.
The Company does not block, impair, degrade, delay or otherwise inhibit access by its customers to lawful content, applications, services or non-harmful devices.
The Company does not impair free expression by actions such as slowing traffic from particular websites or blogs.
The Company does not use or demand “pay-for-priority” or similar arrangements that directly or indirectly favor some traffic over other traffic.
The Company does not prioritize its own content, application, services, or devices, or those of its affiliates.
When talking to an ISP network admin, he told me they tried doing QoS and traffic shaping, but issues always cropped up and customer would complain because of poor performance.
Ah, so you heard from a guy that it's this way. What do you *know*? I've worked on the QoS for about 10 ISPs of various sizes (some as a network engineer, some as the network manager, some as the equimpent manufacturer's support), and you'd have to be pretty dumb to set it up that broken. So the (wrong) word of an idiot admin doesn't trump what I've actually seen in networks with 1000 to 1M+ subscribers (And all sizes between).
Learn to love Alaska
Well, it is a privately own ISP that started back when telegraph was all the rage. They also openly turn down government grants and loans. We're also in a small town with a high unemployment rate and low median income. Even in this horrible situation, they still manage to sell uncapped dedicated symmetrical internet for a fraction of the price of Charter or AT&T. My $45 home connection has magnitudes less jitter and a fraction of latency to the rest of the internet than my job's enterprise 10Gb connection to Charter. They only reason they stick with Charter is they are cheaper for enterprise bandwidth.
Forgive me if I assume they know what they're doing.
I would word this differently.
I would say that ISP's should allocate enough bandwidth for the service they provide. But, of course, if they can avoid doing this (and many times, thanks to monopolies, they can) and save money, they will.
It's rather like medical insurance: companies have no more incentive to provide better service than to save money through simply risk selection by cutting out customers more likely to get sick.
Hence the need for legislation.
Kythe
If, by "sensible", you mean "it makes economic sense that companies will provide crappy, less expensive service when they can get away with it", then I agree with you.
Kythe
There's a large market and a lot of societal benefit to providing crap internet access at an affordable price.
People who want high quality low latency bandwidth can get it, just not cheaply.