Hacker News new | past | comments | ask | show | jobs | submit login
Internet speed guarantees must be realistic, says Ofcom (bbc.com)
199 points by ohjeez on Oct 8, 2017 | hide | past | favorite | 187 comments

It always puzzled me that internet companies can simply say that the speed is "up to" a certain number. Especially since, they will often sell various tiers of speed, proving that they ARE capable of delivering a faster speed.

Imagine paying for 50mb/s internet and only getting 23mb/s. So you upgrade to the 75mb/s plan and now your real speed is 46mb/s. Why couldn't you get that higher speed for the original plan? It seems bizarre that this setup would be legal anywhere in any country.

If you can simply provide "up to" on something, then I wanna sell you a box for ten bucks that contains up to a million dollars!

Edit: Obviously speed depends on a million factors, but I don't know if I've ever experienced even half the speed I was paying for, even when I control both end points.

Imagine paying for 50mb/s internet and only getting 23mb/s. So you upgrade to the 75mb/s plan and now your real speed is 46mb/s.

In reality you'd probably still get 23 after upgrading. The state of New York is suing Spectrum over this issue; customers complained about slow speeds so Spectrum told them to upgrade to a faster plan but their speeds didn't improve at all because the congestion was still there.

They should force the cable companies to automatically discount the monthly cost if, say, 30%+ of the actual use falls more than, say, 10% below the advertised speed.

I mean it makes perfect sense. And oh, what fire it would light under their asses.

Yeah, it'd be perfectly reasonable if it was prorated. So if during a month they managed to provide an average of 70% of the promised speed, I should pay them 70% of the subscription cost.

Average is not a good measure for that, because there are lots of hours during the day where even the worst ISPs don't have any congestion, and still the connection can be unusable during peak hours. The right measure would be the minimum throughput that you achieved, maybe minus a few minutes of congestion per month that are ignored. If they sell you a 100 Mb/s pipe, that should mean that you should essentially never notice that it's not actually reserved 100 Mb/s. People don't buy 100 Mb/s because they want 100 Mb/s at 3 am, they buy 100 Mb/s because they want 100 Mb/s when they need it, which in most cases is during peak load times (which is why those are the peak load times).

I don't think this would work since you can't effectively measure bandwidth available without stressing the system?

You can't do it properly from an endpoint, but the ISPs are already monitoring it internally. They know when and where they have serious congestion.

This is precisely why ISP's should have to operate more transparently as a common carrier and not as a private business milking a cash cow.

If an electric utility company did the same, with 'up-to 240volts' - they'd get their ass sued out of existence. Why can't the gov't establish standards for such things, like a minimum speed (from the IPS's own servers of course, you cannot expect that they control it all).

They would just game it then and underinvest on peering (similar to what happened in the netflix/Comcast scenario). What good is 1gbit guaranteed to your ISP if the entire ISP only has 10mbps to the rest of the world?

Presumably peering congestion would be one of the regulated metrics, just like voltage would be for an electric utility.

Better than having 10mbps to your ISP and your ISP having 10mbps to the rest of the world, if you're paying for 100

Why not? If the ISP provides a server you can download a file from, you can see your speed, the ISP can see the speed. Yes there might be issues with your PC and such, but this can be easily tested.

That would be stressing the system. My point is as a consumer, you can't continuously measure your maximum bandwidth without using it all.

Well you always have Layer 1/2 data. DSL modems, cable, LTE ... will show diagnostic information on which you can estimate some bandwidth, more or less.

That doesn't help you figure out if there's congestion two hops away in the ISPs network due to them oversubscribing their backhaul links.

True. That's what speedtests are for. Ideally you have your own server, so that you can factor out at least one third party.

You are correct in theory.

However, the present system is so bad that it routinely fails at (eg) streaming Netflix even though this uses less than your advertised bandwidth (by a wide margin).

So at present, it would just be a matter of automatic documentation of failures -- and the situation would be much improved if we were up against the limit you propose.

Fair point.

God the internet provider racket could be my least favorite thing about NYC. The infra is terrible, ping is pretty terrible, price rates are terrible, providers are terrible and you probably only have one for your building. Its a real shit storm. There are too many people for the current infra there and I'm sure no one want to foot the bill to improve it.

I had Spectrum when I moved from NYC. I probably got 50/50 out of the 100/100 and during times of high traffic (dinner until 11) it was probably more like 20/20. I called, they said eh, you can pay 250 to get a better connection put in.

I've never seen a cable company offer 100/100, it was 100/10 or 200/20 or 300/20 (mbps) last I checked.

In my experience, the internet available in NYC is way better than most suburban or rural places in the US. 50/50 is actually pretty good, if that's what you were getting - that's like 5 high-definition video streams both up and down.

It's bizarre to me that you think that 50/50 is good. I'm on 1gbps up/down and using slower connections at home is not ideal for me.

1gb/1gb is incredible, and probably top 1% in the world of connections. My office connection is 1gbps/1gbps, and my home connection is 15mbps/2mbps, with no possibility to improve. The highest package available in my city with any provider (not available in my area), and probably across most of the uk is 300mbps/30mbps

It’s more than sufficient to watch Netflix, download things in reasonable times, and browse the web.

Few have 1gbps...

I would classify 1gbps symmetrical as "stupendous". 50/50 would be "good". I'm paying too much for 50/10, and apparently being delivered about 15/5 (per my ISP's speedtest).

The median connection in Melbourne, Australia is around 6/1. I'd count your blessings. Last year my house was only serviced by one ISP and the only connection available was 1.5/0.2 ADSL. Half the modern webpages we go to don't even load at those speeds.

Source? I live in Fitzroy and while I’d agree that it’s generally woeful in Oz that seems a tad low. Experience at multiple rental properties puts it closer to 12/1.

I just got my FTTB NBN last week and I’m getting 94/34. I am over the moon, which will doubtless make some here chuckle. That’s an epic connection for Australia.

(Theoretical max for FTTB VDSL is 100/40.)

My last 4 properties have been 4Mb/s, the aforementioned 1.5Mb/s, a place that had 30Mb/s cable, and now 6Mb/s. All south-eastern suburbs.

I'd say inner suburbs like Fitzroy are probably batting above average. You've got much shorter distances to the DSLAMs there for ADSL2, and more businesses which means more infrastructure.

Obviously the average and median are going to go way up now that they're connecting FTTN and HFC everywhere. From the people I've talked to, the average seems to be around 20-30Mb/s. My parent's house is actually slower now that they've officially installed HFC there (they were getting 30Mb/s on cable, now 20) :/

Woeful is definitely a good word for it. Clusterfuck would be up there too.

Our cable internet is sold as "up to 100mbps", so the joke at our house is that if we ever get 101mbps we will sue them for breach of contract.

Edit: I realize it isn't a good joke.

The New Zealand Commerce Commission got grumpy when ISPs started selling gigabit plans as 1000mbit, as it's not realistically possible for consumers to reach those speeds. So, instead of gigabit, we get "up to 900/400Mbps" or more realistically about 800/400, which, you know what, I can live with.



Are internet packages better in NZ than in Australia? I'm assuming the connection is heavily metered?

The NZ Ultra Fast Broadband project is delivering a lot more FTTP, whereas the Australian National Broadband Network started out delivering FTTP and switched to a primarily FTTN network after an election caused a change of government, much to the annoyance of anyone who was hoping to finally get decent network infrastructure in this country...

I have an unlimited gigabit connection that I pay $120NZD per month for. And for content on good CDNs or hosted in the country I do get near the advertised speed. It took me just under 4 minutes to download the Star Wars Battlefront 2 beta from Origin which had a size of 23.78GB, averaging 90+ MB/s.

I probably use around 5TB of data per month as well.

Nice, do you have also uplink with gigabit?

In Poland I ave 300/30 Mbit (they connected fiber to my house recently) for 50 PLN (it is about 19 NZD, or 13 USD) per month. They offer 600/60 in some cities (where there is more competition), but I don't know the price, I think is is more in 80PLN range. And plan to start also 1 gbit plan (probably with 100 Mbit uplink).

Not the guy you asked, but here the speeds are set by the fibre infrastructure holders, not the ISP. So on residential connections the fastest you'll get is 900/500. I pay $129 NZD (~335 PLN). Not symmetrical, but damn impressive nonetheless.

In Switzerland I have gbit up and down and pay like 90 CHF (92 USD) in a month.

I use the G Suite Google Drive and get nearly full gbit while up and downloading.

I’m about to get symmetric 1gbit from Sonic in SF for the initial price of $40/Mo... after a while it goes up to $60.

Damn, that's not bad at all! For comparison, I pay 40 USD for a 20 Mbit connection with Xfinity here in the US. Oh, and I only get 1 TB of data per month.

I am so glad I moved out of Comcast territory.

Our government has invested a lot more into broadband infrastructure than Australia has. We rolled out FTTN (Fibre to the Node) for improved ADSL/VDSL years ago, and now FTTH is deployed in a lot of places so far. Australia proposed FTTH, then rolled back on that. So yes, we're generally much better for Internet than Australia.

Nope, 100% unlimited. Metered connections were a thing here 5 years ago, nowdays you generally get 80GB or so to start with and can upgrade to unlimited for a little extra. I pay $129 NZD ($91 USD~) per month for my unlimited almost-gigabit connection which is the most expensive residential tier available, and I can pick from any number of ISPs (there's like 5 major ones and plenty of tiny players) as the infrastructure is shared, unlike what you see in the U.S. with the cable companies.

I pay 45€ for 500/250 in France. I get 950/230.

The move to fiber (it ends on my box) was awesome.

If I may balance things a little here : we have "100Mbit" Charter/Spectrum and for the vast majority of the time we get ...120Mbits/s. True, there are times, usually a Sunday evening, when things go topot on their network and QoS goes down the pan, but those times are rare and the problem seems to get resolved quickly.

Measured at the demarc Ethernet segment, L2 throughput including framing overhead, with a single TCP stream to a server I control that is about 1000 miles away.

Yes, I get 70 on a 50 plan from Spectrum. This is less awesome once you realize why they're doing it: in some places they have serious congestion during peak hours so they are giving extra performance during off-peak hours to make the average daily throughput of FCC speed measurements equal 100%. This was uncovered in the New York lawsuit.

Would be an interesting defense against a bandwidth overage charge.

The article is about the UK. Previous OFCOM rules told providers they must advertise only a speed that at least 10% of customers can reach. So yes, people got faster than the advertised speed, about 10% of people. It makes no difference to the whining unfortunately.

The effect is, for say 40MBps VDSL (the service most UK households with "broadband" use, almost always ultimately delivered through a copper telephone cable from BT's Openreach last mile subsidiary) all the ISPs will advertise that as "up to 38Mbps" because at least 10% of their customers can get 38Mbps out of that link. Nobody gets 50Mbps, the link layer is throttled because they're on a 40Mbps product, but plenty of people absolutely will get 40Mbps even though the advert said "up to 38Mbps".

That includes my very tech-savvy ISP which is obliged to say "up to 38Mbps" in big print everywhere even though it explains that you're buying the 40Mbps VDSL service and talks about deep network internals, and even though its customers usually know exactly what speed they'll get (I get 40Mbps) because they're technical people and know how to measure.

Now, suppose you're one of the people getting 30Mbps from the 40Mbps VDSL product. Should you upgrade? There's an "up to 76Mbps" product advertised by your supplier, surely that would be faster? No. All the "up to 76Mbps" product does is change the throttle to 80Mbps, and you aren't impacted by the throttle so the product will simply cost more and achieve nothing.

Now, in fact under those rules an ISP had an incentive NOT to sell you the expensive product, because if they give you the 80Mbps product then you hurt their "up to 76Mbps" averages and make it harder to sell the product.

The result of _that_ change to the market was ISPs would have a headline product advertised heavily e.g. "Guaranteed at least 60Mbps for £24.99 per month" and then if they couldn't meet the guarantee for a particular customer they wouldn't magically fix things to go faster, they would simply say "Ah, you don't qualify for that offer, you can have a refund and the service stops, or pay £9.99 per month for our other product "Guaranteed 10Mbps" which we're meeting.

The ideology behind all this stuff is that our old friend the "invisible hand" will fix it. But er, no. There is actually only one last mile technology involved (for people in most cities there is a competitor, Cable TV cables, but that's a whole separate ball of wax) and it's VDSL offered at speeds set by the wholesaler. It will perform exactly the same whether you buy your "service" from an ISP with a call centre staff in India that uses a purple logo and has a chimp mascot, or you buy "service" from two blokes named James who work out of a rented office in Liverpool, because the "ISP" part is trivial and largely irrelevant except for branding, and the last mile part is a natural monopoly provided always by the exact same engineers, same copper cables, same everything.

There are often bandwidth caps on each tier, which is a different reason to upgrade.

If you consume a lot of video online, the caps can be hit surprisingly quickly.

I found the edit funny though.

What about turning this logic backwards?

I don't need a guarantee. I want per-mbit prices. Like you pay what you get. Limit the rate by saying: I will pay up to 100mbit for example.

Then the can give you less for less money. But this would not be in their interest anymore.

Plans are for marketing purposes only. Broadband ISPs already typically calculate bandwidth as 1-2 mbps per customer.

Sweden has solved this in a pretty good way I think. As an ISP you have to disclose a range (e.g. 12-24 Mbit/s) and pledge to keep that lower limit. In some areas you see instead the range of 1-XX and then you know as a customer that the connection is less reliable.

They’re selling best effort service at a fraction of the price of what dedicated bandwidth costs. We’ve got a 100/100 symmetric line at work. We can pick from several different business Internet providers, but they all cost many multiples of what my symmetric gigabit line at home costs. It’s a different product offered at a different price.

This is not unique to ISPs. You can buy an over-subscribed VPS for a fraction of the cost of a dedicated server.

They’re selling best effort service at a fraction of the price of what dedicated bandwidth costs.

Unfortunately, what they are advertising is something quite different, and the claims made are often wildly inaccurate.

Over-subscription of shared resources for efficiency is one thing, but misleading advertising is something quite different. Frankly, if they get their maths wrong and consequently fail to provide the expected level of service, that's on them anyway.

That's why home services are always advertised as "up to" X mbps.

That's been a sore point for a long time, hence the discussion we're having today. But even with the weasel words, there are misleading practices going on.

For example, if an ISP is offering several levels of service with speeds up to X, Y and Z (in increasing order), it's clearly unexpected from a customer's point of view for an ISP to move them from X to Y or Y to Z if the customer is already getting below the upper bound speed on their current plan and has no prospect of getting a faster speed in return for the presumably higher charge for a "faster" plan.

Likewise, if there is no realistic prospect of a customer achieving anywhere near the X in "up to X", it's misleading for the ISP to offer them service on that basis.

> there are misleading practices going on.

I don't really know how you fix it though. Would selling priority be enough?

Priority 1: "For gaming and other high performance needs. Typical ping: Xms, Typical throughput: Y/Mbps"

Priority 2: "Perfect for causal streaming and general internet usage..."

Priority 3: "Affordable internet and good performance on off-peak hours..."

Just as every other service that uses oversold networks?

The electricity grid is oversold. Do you normally notice that? No!

The flight network is oversold. Do you normally notice that? Not really. And if you do, they have to pay you for failure to perform, cover your costs ...

So, they get to decide whatever bandwidth they want to advertise, possibly they could reserve the right to cancel the contract if the line turns out to be too bad after all (in which case you obviously don't have to pay anything), and then they have to make sure that you never notice that the network is actually oversold by monitoring load and adding capacity to their network as needed to avoid congestion. Up to three minutes of congestion per month is acceptable. Beyond that, they have to reimburse your payment for the month because of failure to perform.

Obviously, it is perfectly fine to advertise different speeds for different times of the day, as long as you don't advertise "1 Gb/s" and then a footnote saying "from 3 am to 4 am, otherwise 10 Mb/s". Or you could advertise burstable service ("100 Mb/s, 10 minutes burstable to 1 Gb/s").

Sure, but if every internet company started advertising their internet packages more transparently as 'best effort, capped at $speed' would you suddenly be happy with your terrible internet?

I don't thing changing the marketing rules will suddenly bring about the sweeping infrastructure changes you imagine.

"best effort, capped at $speed" is not transparent, it's just different wording for "up to $speed". "Best effort" is a completely meaningless term in this context, as it is completely up to the ISP to decide how much infrastructure they build, and thus what speed you can achieve. There is no fundamental reason why an ISP could not provide you full 100 Gb/s duplex 24/7 at any point whatsoever on this planet--it is all a matter of how much money they decide to spend on it, which is not your problem as a customer. So, wording of that kind in effect simply means "we don't owe you anything in particular, we will decide after the fact what you get in exchange for the payments that you owe us."

The point of actually transparent marketing is not to directly improve the connectivity you get, but to enable a market to actually work. If customers cannot know what service they will be getting for their money, then competition cannot possibly work. Imagine all packages in the grocery store were labeled with "up to 1 liter" or "up to 1 kg", and no way to weigh them yourself before you buy them ... how would that possibly allow the better product to win out? What if someone truthfully labeled their product "100 g" ... how could people figure out whether that's the better product than the "up to 1 kg" package right next to it?

Nothing guarantees that better products would actually happen, but at least, if they do get offered, customers would have some chance to actually know, and thus for the better product to gain customers.

> is a completely meaningless term

It's not totally meaningless, and people would understand what it actually means https://en.wikipedia.org/wiki/Best-effort_delivery.

> There is no fundamental reason why an ISP could not provide you full 100 Gb/s duplex 24/7 at any point whatsoever on this planet

Well yeah... with enough money. I guarantee that you wouldn't like looking your internet bill when you have those kinds of guarantees.

> What if someone truthfully labeled their product "100 g"

The internet version of this would be 100Mb guaranteed with an SLA. Or 'business class internet' which few consumers buy because it's expensive. It's not like it isn't available -- it's popular among streamers.

> It's not totally meaningless, and people would understand what it actually means https://en.wikipedia.org/wiki/Best-effort_delivery.

Please don't quote out of context.

Also, as you claim that it is not totally meaningless in this context: Assume an offer of "best effort internet, capped at 100 Mb/s". Please tell me what actual bandwidth I can expect to achieve 90% of the time, and how you derived that value from that offer.

> Well yeah... with enough money. I guarantee that you wouldn't like looking your internet bill when you have those kinds of guarantees.

You are simply missing the point.

> The internet version of this would be 100Mb guaranteed with an SLA. Or 'business class internet' which few consumers buy because it's expensive. It's not like it isn't available -- it's popular among streamers.

No, it's not. There are other possibilities besides "you have a 100% guarantee that 100 Mb/s is available every second of every day" and "if you are lucky, some of your packets may get delivered".

Also, how exactly would a product become more expensive merely because the ISP told you what the product actually is? If an ISP connects up to 10 1 Gb/s links to one 1 Gb/s uplink, say, how exactly does the price of that product become higher if the ISP told you that informartion (or equivalently, that you have a minimum bandwidth of 100 Mb/s available at all times)?!

A good starting point wold be the speed that's available at least 95% of the time.

Right, but then customers are going to be either really disappointed at that number or the price is going to skyrocket when basically every internet package becomes dedicated.

Under your rule there would really be no way to sell 'best effort, capped at X' internet packages.

You can oversell by quite a lot and still meet an advertised speed with 90 or 95% reliability.

If you want to sell 'best effort, no promises' then don't make promises. Or promise something low, like "2mbps" or "2-20mbps".

The real disappointment comes when you can't reach the advertised speed for multiple hours every day. There is much less disappointment in the end if the advertised speed matches a state of moderate, everyday congestion.

I agree that if you simply can’t get the advertised speed, that’s misleading (though, at least on cable or DSL systems it can be very hard to tell a priori what speeds a given chstomer’s line will be capable of). But whenever this discussion comes up, people push back on the very idea of oversubscription or best-effort service.

Home and "business" lines are exactly the same for pretty much every ISP out there. Real difference starts only for other ISPs (or anyone with ISP level bandwidth requirements), still best effort though.

I was under the impression business lines typically have a different contention ratio[1], so you are less at the mercy of "noisy neighbors," as well as offering real SLAs (for whatever good those are).

[1]: https://en.wikipedia.org/wiki/Contention_ratio

The parents are I think talking past each other somewhat, because there are really three kinds of Internet service and you're talking about two of them as if they're the same thing:

1. Consumer/Residential: speed is "up to", throughput is highly asymmetrical, QoS is best effort, reliability is "mostly works", if you are lucky you will get a human if you call for support. If you complain about service typically the response is a shrug. Generally the service is whatever the provider defines it to be and you either buy it or you don't. Often static IP is not available. Cost: $20-100/mo.

2. Business: Sold by the same providers as #1. Much the same service, except (this varies) : dedicated channels on HFC, different traffic shaping, static IP allowed, perhaps even subnets, if you are really lucky reverse DNS is supported. Generally you get a human when you call support. If you complain about service they might do some basic investigation. In the case of our cableco they have different humans knocking on your door for install and trouble shooting (perhaps the more knowledgeable and more senior folks?). Cost: $50-200/mo

ISP upstream/Dedicated/Real Connectivity: The name varies but this what one of the parent posts calls "Business". The service an ISP would get, or a big tech company, etc. You typically get an SLA with specified maximum packet loss rates and latency profile. Throughput is symmetric. Local loop is typically fiber. You can run your own BGP, HSPR etc is supported. IPv6 is always available. You can call the provider's NOC directly and have a conversation about why they're dropping packets in one direction to some obscure peer somewhere in the world. They'll investigate and resolve the problem asap. If you accidentally power off your edge router one of their NOC staff will call your call phone within minutes asking why there's no traffic on your link. Cost: $1000/mo+

Source: I ran an ISP for 10 years.

There is an entire market segment in-between those: http://www.cogentco.com/en/products-and-services/dedicated-i.... For several hundred to several thousand dollars per month, Cogent, Comcast (Metro-E), Light Tower, etc. will sell you dedicated bandwidth to their backbone and an SLA.

You can't expect $50/month services advertised as offering "up to" a certain amount of bandwidth to give the same level of consistency.

you can get ELM (30 /40 MBS) in the UK for £70/month we used it for a stop gap whist our1MBS was provisioned

Where I used to live many years ago. I paid for up to 30. Got about 23. Then I switched to cheaper 10 and still got 23. And told all my friends and they got the same. They had 3 tiers all providing identical service beyond who was willing to pay for 10,30 or 50. Did this for 3+ years.

Actually, this is an interesting topic, because there are court decisions on this in Germany.

So if an ISP has, say, a 25mbps, 50mbps, and 75mbps plan, the 50mbps plan has to provide 25 or more, and the 75mbps plan has to provide 50mbps or more.

I decided last year that I want to move to Germany within 5 years. (For various reasons) Since I've made that decision, I keep finding more reasons to do so.

You know how ADSL (and related tech) works the speed you get depends on the distance from the exchange I don't know how it works in the USA but I ca test my line and at 6k yards it said I should get 3.5 ish which is what I get .

I suspect a lot of BB speed complaints are down to poor wifi installations and local congestion in the wifi bands

See, this is what bothers me. There are people who mistake the speed of their internet with the speed of their own network. It bothers me a lot that when I call about slow speeds that the ISP starts troubleshooting my network. I know not to test on the wifi. (mostly test on the modem itself)

I have a 500/500 line, and get guaranteed that it wont fall 10% below that. Most of the time I am really close to it or even above it. Right now I am seeing an anomaly 420/220. Not going to complain though, I only have this line because the other option was 20/2 for 80% of the price im paying now. Does help when getting or sending big files over for projects. edit: speedtest was giving me a slow server, better results now http://www.speedtest.net/my-result/6690112916

It sort of depends on distance, in reality it depends more on circuit noise and attenuation. If you have a really quiet circuit and high gauge high quality copper you will get a good signal to noise ratio which will give the modem margin to train to.

The curves used to calculate speed over distance make bulk assumptions.

>If you can simply provide "up to" on something, then I wanna sell you a box for ten bucks that contains up to a million dollars!

So, scratchcards then?

The classic excuse I'm receiving from Virgin Broadband at the moment is "we're currently upgrading the network in your area, there may be some speed disruption for several months." I'm paying for 100mbps, a few weeks back I tested it during one bad spell; I was receiving less than 1mbps.

I was struggling to sync my emails, never mind watch Netflix.

Not to mention that right now (and in fact for the past 4-5 days) Virgin Broadband is utterly broken.

I get 38% average packet loss over long time windows, and see 2500ms+ roundtrips to the next hop. If this area was served by Hyperoptic, I would have changed more than two years ago.

If my area was merely served by two tin cans joined by a piece of string I'd have changed years ago.

Australia is appalling for this, the average person paying for 24Mbit ADSL2+ gets a messily 7Mbit (and I thought 24Mbit was bad enough!).

In Melbourne I never experienced sustained speeds of below 9Mbps on ADSL2+ in two locations, normally 11–12Mbps.

But if you want a fast, stable Internet connection, move out into a small country town with a 4G tower (doesn’t need to be NBN). There’s much less congestion. Look around for plans, as most are expensive; I use amaysim’s 50GB/$65/28 days which is about what ADSL2+ would be in the country. Living in Stawell (a few thousand people) I get 20–25/12Mbps with greater reliability than I ever got on ADSL2+. I shall be moving to Navarre soon (~100 people) and get 45/15Mbps there.

If you can live with the data cap, that's great. Otherwise, 50gb is pretty low for work (remote conferences, download some system images, upgrade your mac and you're done).

Another issue in small towns is that the moment a new tower needs to go up, there may be a low of community pushback (radiation! kids!). It's much better if the current one gets an upgrade.

Small towns in Australia tend to be a lot less paranoid about technology than an equivalent sized town in the USA. Comes from being really remote (think 100+Km to the next centre), having to fight for any services at all and a long history of using telecommunication technology to make life better. Think School of the Air using HF radios to provide education to remote children.

Some possibly are. Mount Beauty had a weeks long local Facebook posts about it when wireless NBN was being planned.

Then there are cases like this: https://wangarattachronicle.com.au/2017/08/30/solar-flare-up...

Yeah I can't imagine using less than 500GB~ month myself, wireless isn't the answer as it does not scale at all well.

> It always puzzled me that internet companies can simply say that the speed is "up to" a certain number.

I always imagine it like this:

    Them: "This plan gives you up to 23mb/s."
    Me: "So you're saying you guarantee I will not get more than 23mb/s?"
    Them: "Right."
    Me: "OK. I'll pay you up to $40 a month for that."

Same for "Unlimited" when it actually has clear limits. Or camera manufacturers sell "Weather resistant" equipment but never specify what it means exactly and will void your warranty if there is water damage.

Having their cake and eating it too!

Tangentially related: We used to have a bunch of "$2 Shop" stores - I think other countries have the equivalent "dollar store", "pound shop" etc. Due to inflation they've all now become "$2+ Dollar" shops or "$2 and up" etc.

In reality of course they still sell the same sort of cheap stuff but it's funny how the name has gone from a shop that guarantees cheap prices, to one that guarantees not to be really cheap, since any price is now valid except ones below $2.

It's because the law hasn't yet caught up with their spurious marketing claims.

It would be interesting to pay in x€/Mbps (up to y Mbps). This is how I buy potatoes so this is hardly a revolutionary novelty.

I'm aware a lot of people have this problem, but here in a rural area we are paying for 6 Mbps DSL and usually get 6.25 to 6.5.

They can. I often had 11mbps on my 10mpbs cable line.

As someone with a company in the ISP space, this is awesome news.

One of the biggest problems in marketing that our team has come across, particularly in consumer broadband, is that as a company attempting to be open and honest about speeds and pricing, it can be hard to compete with the older players using bamboozling pricing and inflated speed claims to trick consumers thinking that they are giving a better deal.

These restrictions will level the playing field in the right direction for a more informed and better served consumer.

At a time of some very bad calls around internet legislation in the UK, this is finally a decision I think we can all applaud.

Yes, that sounds like a very good idea for both sides. It's annoying as hell when as an informed consumer, you have essentially no chance to figure out what the products even are that are competing for your money. How the hell am I supposed to decide between products that only guarantee that they won't provide more than a certain level of service with no lower limit? Just imagine supermarkets filled with "up to" products ... WTF?

It's a disgustingly unhealthy market when the only possible way to find the best product for you is to buy service from each available option for over a year (in order to get the true, not stupid introductory price)

More transparency from ISPs would be amazing.

Despite their flaws, one thing I love about my ISP is that they openly publish the utilization MRTG graphs for their local fiber connection nodes.

So you can see where I live, they're starting to get close to saturation http://qos.plala.or.jp/traffic/flets/kagoshima.html but in other parts of the country they've already started to peak out http://qos.plala.or.jp/traffic/flets/chiba.html

In a weird way, this is the same problem that banks face.

Fractional reserve banking is actually in society's best interests: it's fine if the bank takes $100 in deposits, then loans out $80 or $90, rather than just having the money sitting in a vault doing nothing.

In the same way, it's probably better for the overall efficiency of communications for ISPs to oversell capacity and then allow bursting on top, rather than not overselling and wasting bandwidth when there's no congestion.

They face the same problem though: banks have an incentive to lower reserves until the smallest shock will bankrupt them, and ISPs make more money by overselling past what is wise or reasonable.

Same solution: for banks we regulate margin requirements, and for ISPs we should regulate how much they can oversell.

That's not how it works. Fractional reserve of 10% means that $100 of deposits could theoretically turns into over $900 (The central bank loans $100 to you, you can then loan $90 to Pete, Pete can then loan $81 to Sam, Sam loans $73 to Gill, etc.). Here's an example of how it works (for 20%):


That's why people talk about banks being too big too fail now. If one bank fails, where do the loans go? No-one can pay off the loans immediately as they all only have 10% of the assets. So we can't call in all the loans. But another bank can't afford to buy the failed banks assets as they would have to up their reserve, but that money doesn't exist any more as the original bank is bust.

One large bank failing would mean a huge chunk of theoretical money would just, poof, disappear.

That's why some people say it's fraud, etc. Not that I'm agreeing one way or another, but $100 of deposits is worth much more than $90 in the system.

Here's a table showing the total money that could theoretically enter the system for each fractional reserve percentage:


Well, first of all, banks cannot simply claim that they don't have all your money available and give you back only half instead. Once they reach a stage where that is their problem, they are bankrupt, and you get to sell their assets to recover your money.

I think it's much simpler in the case of ISPs, because there is no risk involved for the customer: You don't need to regulate how much they can oversell, but rather simply that the customer must not ever notice that the network is oversold.

I mean, they can during a run on the bank without declaring bankruptcy.

No, they can't.

First of all, only paying out part of the balance at once is very different from paying out part of the balance instead of the full balance. A bank can not unilaterally declare that their 100 dollar debt to you is paid by giving you 50 dollars. If they pay out 50 dollars, they still owe you 50 dollars.

But more importantly: No, they can not even unilaterally delay the payout. If you have demand deposits of a million bucks with a bank, that means that you can at any point demand they pay out a million bucks, and when you demand that, they are legally required to do so, and if they can't fulfill a legal obligation to pay, that is exactly the definition of bankruptcy. If they fail to immediately declare bankcruptcy, the management is on the hook for any losses due to that failure of theirs.

Banks might limit what their ATMs pay out per person per day or something, say, but that does not change anything about their obligation to fulfill any requests to pay out your balance immediately. Just because the ATM doesn't allow you to demand a payout of a million dollars, doesn't mean they have any legal option to refuse if you go to the cashier and express your demand that they pay out a million dollars right now. At best they might not have a million dollars on site, in which case they possibly might be allowed to have you wait for however long it takes to move the cash from the central bank to the branch that you are at, but that is certainly not more than a day.

You're right that the bank cannot unilaterally prevent a payout.

Practically speaking though, what usually happens during a potential bank run is that governments declare a bank holiday, preventing customers from withdrawing funds.

That buys the government (and the bank) time to get their ducks in a row, line up bailouts, etc.

Things get a LOT worse if the bank is actually allowed to go under. Letting Bear Stearns and Lehman collapse was, in retrospect, probably a mistake that made the eventual bailout need to be larger.

Well, that really depends on the bank, though. In the case of smaller/local banks, chances are that it will simply go bankrupt to be wound down and deposit insurance will pay out the balance (or as much of it as is insured) within a couple days.

But also, bankruptcy does not necessarily imply that the bank collapses: One way to resolve a bankruptcy is to obtain funds, possibly in the form of a bailout, that allow any demands to be fulfilled. Rather, bankruptcy is exactly the thing that allows the bank to refuse payouts even when it's not a bank holiday. The point is that once a bank has declared bankruptcy, it is not allowed to pay out to anyone anymore without oversight, because the bankruptcy proceedings are supposed to make sure that everyone that the bank is indebted to gets a fair share of the remaining assets--or possibly to find a solution that allows the bank to resume operations by accepting a later payout for some of its debts. The important point in all of this is that in any case, none of this can be used directly to increase the equity of the existing owners. If the bank has to obtain loans in order to fulfill its obligations, the interest on that is paid from equity. If there is no equity, the bank changes ownership (like, if the state bails out the bank, the state gets to own the bank). So, even if the bank might refuse to pay out under certain circumstances, the processes that get triggered by that have as their primary goal to make those payouts happen as soon as possible, not to allow the bank's owners to keep anything.

Many factors can influence browsing speeds, he said. These include

...and not one mention of the server at the other end. Whenever discussion of Internet connection speeds comes up, I always ask myself, "speed between what?" Even if you have a 1Gbps connection to your ISP, you're not going to get anywhere near that if the other end is on a different continent, a small site behind the Great Firewall of China, or a heavily overloaded server.

Thus, in some sense, ISPs have always been advertising maximum speeds, and even those expensive services with SLAs etc. very carefully define to what part of the connection such guarantees apply.

But when I can't even get that speed under favorable conditions then it is dishonest to advertise that speed. If you can't even get that speed when connecting to the ISP's own server a few miles away then you aren't getting what you pay for.

When car companies advertise the fuel efficiency of their cars those are realistic numbers for regular driving because there is a standardized testing protocol. You can trust the number because it is regulated, so why can't we hold ISPs to the same standard?

If you can't even get that speed when connecting to the ISP's own server a few miles away then you aren't getting what you pay for.

That I agree with --- the connection from your house to your ISP should always be at advertised speed, but beyond that it's basically impossible to guarantee anything.

Actually it needs to be /through/ the ISP.

Them selling you a fast speed is worthless if their peering points are saturated beyond any sane value.

How is it impossible for an ISP to guarantee bandwidth within their own network, including to/from interconnection points/exchanges?!

Maybe not at every point in time (if you accept some overbooking), but that doesn't mean that you can't at least guarantee bandwidth for > 99% of the time.

The ISP's advertised bandwidth should be available essentially always, with rare exceptions for unusually high levels of traffic, between your connection point and some exchange location where other ISPs can interconnect with your ISP without paying your ISP any money. The ISP doesn't have to guarantee that every server is reachable at advertised speeds, but they should guarantee that if ever anyone wants to send packets in your direction at the speed of your link, they should be able to either bury their own lines or buy transit from a market of transit providers to deliver the packets to some point in their network, and that those will be delivered without major packet loss at full rate.

> The ISP's advertised bandwidth should be available essentially always, with rare exceptions for unusually high levels of traffic,

You can buy service like this, and you pay for it. At work, we have a 100 mbps line from Cogent, with dedicated bandwidth to their backbone. It's something like $500-700/month. At home, I've got an "up to 940 mbps" line from Verizon, that's usually about 900 mbps but sometimes can fall down to 500 or so. (It's 16 users sharing a GPON node so only 75 mbps per user if everyone is downloading at the same time). But it's only $90/month...

"essentially always" != "dedicated"

Also, how is "but we couldn't deliver that product at that price" in any way a justification for then simply delivering a different product under the same advertisement? If you cannot actually provide 940 Mb/s for 90 $/month, that's fine ... but then don't advertise it as such.

There is nothing wrong with overselling--but it should be the responsibility of the ISP to hide it. From what you write, it seems like what they are actually selling you would fairly be advertised as a 500 Mb/s link. There is no need to actually have reserved bandwidth for each customer, but the moment the ISP saturates a link for more than a few minutes per month, it should be their responsibility to add capacity, or to reduce advertised bandwidths. I as a customer cannot just go "paying the full price this month is too expensive, I'll only pay half", so why should it be an excuse for the ISP to say "building infrastructure to provide advertised services is too expensive, we'll only deliver half"!?

> delivering a different product under the same advertisement?

It's a different product delivered under a different advertisement. Business lines are advertised as "dedicated" X mbps. Consumer lines are advertised as "up to" X mbps. "Up to" does not mean "essentially always." If I see the Microsoft Store is having a sale of "up to $400 off on Surface Pros," I don't expect a $400 discount on "essentially all" the models--I know it's the high water mark.

Compare to how other over-subscribed systems are advertised. To pick a random example: https://www.serverhub.com/vps/ssd-cached. VPS servers are typically advertised as "4 cores." Are VPS providers ripping people off because you can't expect a $20/month VPS to get you 4 cores of throughput "essentially always?"

> Consumer lines are advertised as "up to" X mbps. "Up to" does not mean "essentially always."

If it doesn't mean "essentially always", what does it possibly mean then? Sure, the literal reading suggests that anything that doesn't exceed X Mb/s is acceptable ... so not connecting the link at all is acceptable then, I suppose? If not, why not?

> If I see the Microsoft Store is having a sale of "up to $400 off on Surface Pros," I don't expect a $400 discount on "essentially all" the models--I know it's the high water mark.

Which is a nonsensical analogy, as one of these is about getting your attention (in order to then tell you the actual specific price of an actual specific product that you could buy if you wanted) while the other is about getting you to make a contract only to then have the other side decide what they actually want to deliver in exchange for your payments.

> Are VPS providers ripping people off because you can't expect a $20/month VPS to get you 4 cores of throughput "essentially always?"


How the fuck is it my problem that they advertise something that they cannot deliver?

If a provider cannot provide 4 cores, then they cannot advertise 4 cores, how could that possibly not be obvious? If you promise something, it is obviously no excuse to then claim "but you should have known that that was unrealistic!" It's your fucking responsibility if you make a promise to make sure you can fulfill it, and if you can't, to not make a promise.

How is a market supposed to work where advertisements are just non-binding suggestions? How the fuck am I supposed to select a provider for a given use/load if every provider can just excuse themselves from providing the advertised service by claiming that I should have known better than to believe their advertising? How should I even possibly know what prices are unrealistic if all prices are made-up bullshit?

Also, why is that only an option for the provider? Can I also just pay half because it should have been obvious that I don't really want to pay that much?

In Germany, 1&1 is renting out 100Gbps lines with dedicated fiber (as in, you have an entirely separate circuit) directly to the IX for 7200€/month.

If you're paying 500-700$/month for a 100mbps line, then something seems very wrong.

Without details it's hard to compare. Does that include initial construction, transit, etc? Is that from a data center, or will they run it to a telecom closest in our 1960s office building?

Also, the price for these lines generally doesn't scale linearly: https://docs.google.com/spreadsheets/d/1xF2Yo-3J2EUjLlsXFYE4... (a quote some random person posted).

That's excluding construction (if you rent 2+ years, it's inckuding construction), including transit, and goes to any building in any of the cities where they offer it. Which is about 3 cities right now, including Kiel.

Overselling of the last mile infrastructure. If you want to see what a not-oversold last mile connection (with a real minimum speed SLA) would cost, take a look at the “for business” pricing and weep.

Most people’s usage is bursty, so the present oversold state is probably the least bad option compared to having a fixed <1Mbps line speed or paying thousands per month for a truly dedicated 50-100Mbps.

It WAS bursty when most people were browsing but it's continuous if they've streaming these days.

Yeah but not much - stream Amazon Prime UHD pretend 4k and consume.. 18mbps. You can support 1k homes doing that on a 10G pon or VDSL build

The for business prices are mostly to do with sla's for repair. If you move from 3 days to 3 hours the cost isn't 32x it's 300x (and more)

Which is an answer to my question how exactly?

They can and do provide guaranteed speeds; it costs a lot more. They can't provide guarantees (or 99% guarantees) of existing advertised speeds at anywhere near the existing price structure because the economics are based on overselling.

It really shouldn't cost /that/ much more. If there were actual competition in a sane market the price would be practically whatever share of the electricity for powering the network infrastructure along the path plus some part replacement overhead and a small profit margin.

That's why the natural monopoly of the last mile should be owned by the municipality (the people in the area) and competition should begin at designated peering points where backbone access can be brokered in an actually competitive way.

Why do you think it shouldn't cost a lot more? A consumer-grade fiber service will put 16-32 passive fiber strands on a single GPON node and some multiple of that on each 10 gbps network port in the intermediate routers that eventually dump into the backbone. If you're selling dedicated bandwidth, you need active fiber out to the customer site, and your brutally expensive core routers can only have 5 customers per port instead of 50.

Moreover, there is actually quite a bit of competition in the dedicated bandwidth business line space. We were recently shopping around our contract at work and had at least 3-4 options IIRC.

Having the municipality own the infrastructure out to the peering point is a terrible idea. There's a lot of very expensive hardware between the fiber strand going to the customer site and the peering point.

Well, then they have to advertise the bandwidth that they can actually guarantee for 99%?! What is the justification for advertising speeds that you can expect to not actually achieve?

I mean, yes, it's obvious that overselling is in the interest of customers, and as such customers should accept that under more or less exceptional circumstances the service does not reach the advertised performance, so you maybe shouldn't use it for stuff that reacts catastrophically when bandwidth requirements aren't being met, fine. But how is it anything but deception to advertise bandwidth that the user can as a matter of fact not expect to be available under normal circumstances?

For comparison, take electricity: My connection to the grid might be built to be able to supply 40 kW, say. Now, if everyone were to actually pull 40 kW from their connection at the same time, that absolutely would not work--but that does not mean that electricity supply is regularly insufficient for my needs. Except for very rare exceptions, everyone can always pull as much power as they currently need. There is no brownout when everyone is cooking in the evening (or whatever). The capacity of the network and generators is also oversold, but utilities still take care to ensure statistically that situations where the actual instantaneous demand is not being met are few and far between. I don't see how the same could not be done with IP connectivity--and how advertising anything for which this cannot be expected to work out is anything but simple deception. If you cannot supply 1 MW of electricity at some point in your network ... how does that legitimize advertising 1 MW connections and then simply lowering the voltage when the customer tries to actually use it?

In short: Overselling is fine, as long as your customer doesn't notice, except very rarely under exceptional circumstances. If some product doesn't work under those circumstances, you'll have to raise the price or lower the advertised bandwidth.

How else am I, as a customer, supposed to distinguish an offer that effectively is going to deliver only 50 Mb/s from one that actually does 100 Mb/s during peak times, if both are allowed to advertise as "up to 200 Mb/s"? Where is the incentive for ISPs going to come from when there is no limit on how much they can oversubscribe?

"I don't see how the same could not be done with IP connectivity"

But it is exactly how it's done. The only ISPs that neglect even that are large evil monopolies, but since they are monopolies they can get away with a lot of things and ask 20x the price.

The other thing people don't seem to understand, is that guaranteed bandwidth is not about speed.

Well, yes, good ISPs do just that, sure, I guess my point is that anything else should just outright be rejected as obviously unacceptable, and it should thus be the legal responsibility of all ISPs and thus breach of contract if customers do actually notice that the network is oversold.

No, no one provides guaranteed speeds. Guarantees you are talking about that ISPs provide are only for you being able to saturate the link 100% of the time, which you shouldn't do on home or business connections.

> How is it impossible for an ISP to guarantee bandwidth within their own network

It isn't impossible, just too costly to the typical residential customer who expects to pay <<$100/mo.

Sigh I sugest you look at how last mile is delivered every pair running to the exchange / CO is diferent in terms of what speed you can deliver and thats before you factor in contention

Regulations don't fall out of the sky. Ban lobbying and stop corporations from financing political parties. That would begin to solve this I guarantee it. But so long as your ISP has the ear of lawmakers you're SOL.

We're going to have to act if we want to live in a different world.

MM so you mean stop Rupert Murdoch throwing a hissy fit and bully his direct competitors then - BTW you do realise that OFCOM was set up to keep him happy (source direct quote from the then chair of an all parliamentary group on telecoms )

An ISP should be able to provide a max throughput from your house to anywhere on the internet, where max throughput is defined as the size of the bottleneck controlled by the ISP. Of course you won't get 50mbps to a server with 2mbps ingress, but you should know when your ISP is the bottleneck.

That's actually not a problem. You can always connect to 10-20 good servers (like linux distribution ones) and fully saturate your domestic connection.

> even those expensive services with SLAs etc. very carefully define to what part of the connection such guarantees apply.

That can be done with consumer broadband as well. The problem with current marketing practices (at least in this neck of the woods) is that you can have a defective line on the ISP's side of demarkation box or be renting defective equipment from them, and they will still try to deny responsibility. Then you have things like congestion. It may be mostly outside of the control of the ISP, yet it can still be estimated in order to ensure that marketing claims are in line with what is being provided.

The list of reasons provided by the article just reeks of blaming the consumer. Sure, the ISP should not be responsible for the consumer's poorly configured networks or computers. On the other hand, they should be accountable for problems on their end or misrepresentation of service levels.

You're probably not going to get 1Gbps from a distant server even if that server is not overloaded.

I run some traffic heavy servers and even while my servers can be perfectly capable of pushing more, one of the links between me and you is congested so we're both getting shafted. Cogent is especially guilty of this.

Basically the entire Internet is a mess, last mile is a problem, but so is all the major interchange points between carriers.

The standard is servers in large data centers with effectively unlimited bandwidth speed compared to residential customers, like netflix, google, amazon, steam, microsoft, apple, etc.

In fact these people distribute content via deep caching and multicast provided by the ISP's and the CDN's - all those people (apart from Apple) have distributed infrastructures.

Lot of people in here moaning about cable speed problems caused mostly by congestion.

This article is about the UK, where most people with broadband Internet get it via their telephone line, ie VDSL, and so aren't subject to congestion in the last mile.

In most cases this means their "speed" is a property of the VDSL standard, the length and quality of copper cable in the ground between them and the VDSL node (a relatively new bigger green street cabinet for most of them, vaguely near their older smaller "BT" cabinet).

VDSL works basically the same way as your ancient "analogue" telephone modem, except with way, way higher frequencies and more sophisticated encoding because it wasn't designed to cost $10 and work over actual telephone calls, just make it for about a mile down a copper cable to a $$$$ specialist box.

Short, high quality telephone cables can easily support the same sort of bandwidth you regularly see from say, 100baseT, ie 100Mbps. But very long, poor quality ones will do much worse, VDSL needs to tweak things to give those poor souls some sort of service, while also delivering a really good service to people who are closer, as automatically as possible. Higher frequencies get smushed worse over distance, so the closer people get a wider _band_ of frequencies, more literal _bandwidth_. The consumer equipment and the big green box in the street negotiate how best to get a signal between them, then if necessary the big green box throttles this down to an agreed speed, the actual "maximum" which may be higher than the advertised "Up to..." speed in countries where a regulator demanded such a pointless change to advertising.

The end result is usually if you live 100 metres from the cabinet, you get nice fast broadband and it doesn't matter which ISP you pick. And if you live two miles from the cabinet you get shitty ADSL at 1Mbps and can't watch Netflix and again, it still doesn't matter which ISP you pick.

So in a way the irony is that ISPs advertise speed at all. Speed has almost nothing to do with your ISP if you're using VDSL, which most UK households are. Imagine if car manufacturers started advertising the quality of roads you can drive on. It's the same roads for all manufacturers - who cares?

Distance from pieces of the infrastructure also doesn't always help, I live 100m from the cabinet, and 500m from an exchange in a large development, but they've not managed to lay fibre apparently (guessing investment in Zone 1 London isn't the best political/PR decision for Openreach...), so it's only ADSL "up to 16Mb", but rarely top 2-3.

> if you're using VDSL, which most UK households are

It's been a while since I was in the UK internet industry but I think most of them are using ADSL with VDSL generally only used for fibre or "business class" connections.

Any broadband package over "up to 20Mbit" will be VDSL. I'd say that's a majority, nowdays.


> a little over half of the country still subscribe to the much slower and less reliable ADSL (up to 8-20Mbps) based pure copper line services.

The big lie is that connection speed within the ISP means nothing. Most of the time when I experience packet loss / congestion you can clearly see it's in the connection from the ISP to other providers, and some providers are cheap you will suffer high packet loss and congestion exiting their network. The advertised speed is meaningless when 1 in every 30 packets is dropped. Just small packet loss is enough to slow TCP

Well, packet loss here really is a bandwidth problem, though. The packet loss happens because some link is saturated and packets are arriving faster than they can be transmitted, and TCP simply reacts by adjusting the sending rate so as to not exceed the available bandwidth at the bottleneck.

Still true that the problem often is more the peering than internal infrastructure, yeah.

I used to love speedtest.net, but since finding testmy.net which keeps a log of your previous tests, I haven't gone back.

I typically receive the correct up/down ratio, though intermittent issues may lower it for a day or less. Though various ISPs over the years. It would be nice to have a consistent as-advertised speed though if you happen to live in an area where that isn't happening.

Speedtest.net also logs your previous tests, but you have to create an account/login to preserve them across cookie expiries http://beta.speedtest.net/results

Eh, I guess this is US only?

I am getting 40/30 here in London whereas speedtest.net gives me 900/900.

That is a massive difference.

At some point certain ISPs were disabling throttling specifically for speedtest.net which is why netflix made fast.com. Perhaps thats whats happening with testmy.net?

I just tried, gave me 240/40 as result. (speedtest.net gives me 510/460

Looking at their servers... okay not so many. But shouldnt the speeds be more correct, and only the ping higher for further away places?

Further away places also mean increased chance of hopping through intermediate networks which could introduce their own bottlenecks.

testmy.net seems to be mainly USA based, but for those outside the US, you can set the default server to one located much nearer your own location. They offer a few international servers and London, UK is one of them, plus alternatives in EU, Asia and Australia

My favourite questionable practice is from Hyperoptic. They advertise gigabit broadband, and this is accurate.

But the wifi hotspot they provide is only capable of a small fraction of that (theoretically 130mbit, practically 70mbit), but you can get the full gigabit if you use ethernet. This discrepancy isn't mentioned at any point until you start digging around in the FAQs or in some forums around the internet:


But I plugged in a proper wifi hotspot (one capable of at least gigabit) and I got the full advertised speed.

I believe this is an intentional trick to avoid saturation of their network.

A few years ago I subscribed to a 8 MB service with a major ISP (Virgin). I was getting around 500k. Virgin insisted it was an issue with my router/my line/my computer/everything under the sun and nothing to do with them. When I pointed out that I had signed up for a 8 MB deal, they said that legally they did not have to do anything unless my performance went below 300k! I cancelled the contract and was forced to pay an early termination penalty.

I went with a smaller ISP (Fast.co.uk - those guys are brilliant!) and immediately started getting around 6 MB. Same computer, same router and same lines as before.

It boggled my mind that a big ISP could get away with this borderline fraudulent behaviour.

I’m not really bothered about speed. Just reliability. If my home broadband could actually stay up then I’d be able to work from home without tethering to a phone in order to get a reliable connection.

Whenever my internet is sluggish enough for me to call Comcast, they always blame the fact that I'm using wifi — and suggest I switch to ethernet to get higher speeds.

I tell them that all of my devices are newer and therefore do not have ethernet. And I point out that all of the service techs they send out do not have ethernet-equipped devices either.

This generally does not convince them. Sometimes they are convinced if I can do a transfer over my network (Time Machine backups, etc.) to show that the wifi is blazing fast, and that the bottleneck is their connection.

You could run the test from their modem, and they would still blame the device you are using to start the test.

These people on the phone can only work off their check lists, once they hear something that gives them an easy out they will start blaming you.

Do you also complain that your SSD is slow when you connect it over USB 1.1?

Not everyone's WiFi is that bad. I get iperf3 results of around 900 Mbit/s over my 802.11ac WiFi. When my internet connection drops to 100 Mbit/s, the WiFi isn't the bottleneck.

When I'm using my SSD over USB 1.1, and the SSD is the bottleneck, then yes, I'd complain.

Just like how I'm right now using a spinning HDD over public internet via networked storage, and the bottleneck is... the storage driver peaking all CPUs at max. (spoiler: it's portworx). And obviously I am complaining, too.

Not sure how you think this is at all analogous. My internet connection is 12 Mb/s, which could be handled by any recent wifi protocol. Maybe get the facts first next time before jumping to unwarranted conclusions.

It's not just about what protocol you use, physical factors also come into play. If you're in an apartment, how noisy are your neighbors? If you're in a villa, how far away are you, and how many walls are there in between?

Yes, Wi-Fi can have pretty decent bandwidth in the optimal case, but there are far too many uncontrollable variables for it to be possible to guarantee anything with it. It also makes perfect sense to want to exclude it as a variable before making a more thorough investigation.

If you bought a computer without Ethernet then that's on you, not your ISP. Fashion is not an excuse.

Did you miss the part where I said it can move files internally perfectly fine? There is nothing wrong with the wifi, and it's bizarre to see someone on HN make such unjustified assumptions.

Does it make sense to "exclude it as a variable before making a more thorough investigation"? Yes. But if Comcast sends out techs who do not have ethernet-equipped hardware (as I pointed out in my original post), then it's silly of them to expect customers to all have ethernet-equipped hardware.

> Fashion is not an excuse

Oversimplification is not an excuse either.

If my SSD was unable to saturate a USB 1.1 connection, I would absolutely complain.

You know, just maybe they weren't using 802.11b...

It would not convince me either. Usually™ Wlan is very slow, partially due to the equipment ISPs sell. Having a new device does not mean you can't use ethernet. Every PC has it. Buying a notebook without it seems backwards, in either case you can use a USB 3.0 ethernet adapter.

This is where it's useful to have a router that will run a test with speedtest.net from the router.

Or just get a USB 3 to Ethernet adapter (several in the $11-15 range on Amazon), which gives you more flexibility to run different kinds of speed tests and troubleshooting, and can also come in handy during travel. Even works on iOS or Android when combined with a different adapter.

Speedtest.net is a horrible way to test your internet speeds. The isps intentionally prioritize traffic to them because it makes them look good, and speedtest doesn't offer any actual services that could create legitimate traffic. At the very least use fast.com, which is run by Netflix and will show you what kind of performance you'll get to a real site. Ideally you'd test against multiple sources though, especially now that Netflix has started giving ransom money to some of the larger residential isps.

fast.com is a great way to evaluate your QoS to the Netflix CDN but in my experience it isn't a good way to test your general Internet-wide throughput. speedtest.net (with a sensibly chosen server -- not the default which will tend to be your ISPs) is better. Better still test to a server you control yourself.

Once again... speedtest.net is white listed by every ISP out there to give maximum speeds and bypass any throttling they have in place and typically gets placed in a prioritized queue. It's just slightly better than useless and will tell you nothing about your actual speeds during network congestion..

No offence ment but if you have drunk so deep of the apple hype and abandoned ethernet you have to take some responsibility here.

Unless of course you are a wireless CCNA and have set up your wifi using professional standard gear.

Any guarantee is unrealistic, depending on who you ask. There are so many points of potential slow downs that it is. Not realistic. To say a minimum.

The problem isn't a lack of a guarantee. The problem is _never_ getting the service as advertised.

If I buy a 50 Mbps plan, and I get 50 Mbps 60% of the time and between 20-30 Mbps 40% of the time, I could live with that.

My problem is that has never been my experience with major cable providers.

I pay for a 50 Mbps plan and get 20-30 Mbps 50% of the time and 10-20 Mbps 50% of the time. That's not simply a lack of a guarantee. That's a complete failure to provide the service that was advertised.

Again (as above) I feel I need to speak up that we have a bigCableCo connection (Charter) that delivers 1.2x the sold throughput. From talking to the local guys there, they have a policy to over-provision connections by about 10% so they can be sure a speed test will deliver what the customer paid for.

When I ran an ISP it was common for customers to run a speed test, see a speed less than they had paid for, and immediately call us. Often the reason was their router or their WiFi, or some flakyness in the speed test site, but sometimes it was our network. We always investigated and resolved the problem to restore their expected speed.

This being the case, I find it quite hard to believe that there are ISPs who as a matter of course just don't ever deliver the advertised speed. They would be receiving constant calls from disgruntled customers, which surely would cost them more to answer than it would cost to fix the network??

You can easily (through shoddy, cheap outsourcing body shops in India, Phillipines, etc.) get support costs low enough that just taking the calls is cheaper than fixing things, at least until the end of the financial quarter, and that's what really matters to management. I say this from experience in the web hosting industry, but I don't see any reason it wouldn't apply to ISPs.

No the problem is unethical companies that sell 6mb dsl from a DSLAM fed by 2x t1s, or that refuse to upgrade their peering links because they want their peers to pay them too.


Good luck imposing anything in US now, with the current corrupted head of FCC.

Ofcom is a regulatory agency in the UK, this article has nothing to do with the FCC and the US.

Please don't post uninformative flamebait.


As the article points out the real speed depends on a lot of factors. Also, not all servers can serve you at your max speed. Currently what would be a decent speed, meaning you experience no lag...20 mb, 50mb ? Has anyone gone from 50mb to a lot more and felt a huge difference ?

You will notice a difference if you live in a house with multiple people streaming at once. The most typical use case is a big family that wants full HD on their tablets plus a 4k stream in the living room. That requires a 100 mbps connection due to speeds slowing down in the evening (part of what the article is about), and needing a bit of extra speed to prevent buffering.

I noticed a huge difference and it’s just me and my server (and two cats but I don’t think they’re using Netflix...).

Normal day to day sites, no real difference. Loading up a steam game or the like, huuuuuge improvement. Also no talk about upload speeds which are an even bigger part of making the web faster.

Make sense, regular sites are usually under 10mb (2+mb average) https://www.wired.com/2016/04/average-webpage-now-size-origi... so load times are essentially the same if you have a 10mb connection or 1000mb

Big issue is wifi; if you live 50m from your neighbours, no problem, but most people are sharing the wifi bearer with 10's of others.

Equipment too. Using whatever junk your isp gives you isn’t going to do too well. Better to spend a few hundred bucks on something like a UniFi setup.

browsing, no. steam, yes

going from 20 > 500 (and 2>500 up)

The upload makes it really nice, with all the cloud storage its really easy to back up a device in the cloud when I am wiping it. Faster than usb transfer.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact