Hacker News new | past | comments | ask | show | jobs | submit login
How a group of neighbors created their own Internet service (arstechnica.com)
430 points by Deinos on Nov 1, 2015 | hide | past | web | favorite | 139 comments

> “We monitor all the connections and if someone is using a lot of bandwidth for a long period of time, we talk to them and figure out what they are doing,” Sutton said. “Often times it's people watching Netflix and then falling asleep and then it keep auto-playing things all night long.”

This is the kind of statement that makes me long for a faceless, impersonal ISP. As much as I'm impressed by a neighborhood banding together and coming up with a solution, the idea of a neighbor trying to "figure out" usage or "talk to" customers about it is horrifying.

Hopefully they have a mechanism in place to allow everyone to burst to available bandwidth, but to throttle people to a sustainable proportion otherwise. Such a mechanism seems both necessary and sufficient, and given the professional equipment they're using, it seems quite likely to be available.

When Comcast sends bandwidth usage nastygrams, people get up in arms about it; we talk about "network neutrality", and that ISPs should remain "dumb pipes".

While the ability to introspect traffic at all is a bug that needs fixing in client and server software, to combat surveillance, at least with an ISP I'm reasonably confident that only 1) a trusted subset of ISP staff and 2) the government (hopefully with a warrant) have access.

By contrast, would you want your neighborhood association looking at your ISP logs? Or anyone you know personally? And making it their business how much bandwidth you use and for what purpose?

(On a separate note, I wonder how much Netflix's and YouTube's CDN boxes cost, given that Netflix and YouTube tend to subsidize them.)

> On a separate note, I wonder how much Netflix's and YouTube's CDN boxes cost, given that Netflix and YouTube tend to subsidize them

I worked at a large ISP for 4 years, and for us, both the YouTube and Netflix CDN boxes were free.

However, the boxes consume a massive amount of traffic to keep themselves in sync. For us, YouTube wasn't worth having until we had over 15,000 customers, and we still don't have a NetFlix box because it syncs the entire NetFlix catalog and all changes, every 24 hours, so they actually use less traffic not having one.

> By contrast, would you want your neighborhood association looking at your ISP logs?

Maybe you misinterpreted? He said:

> " ... we talk to them and figure out what they are doing."

He didn't say, "we transparently force all traffic through a proxy and log what sites everyone is visiting".

I work for an ISP and we have some customers (both business and residential) that are completely clueless about what's going across their network and out to the Internet.

"We monitor all the connections" could mean many things. More to the point, they have the capability to do so; you don't need a proxy for that, just root on the gateway box and a mapping from IP addresses to users. Or, for that matter, DNS logs if they have a local DNS server. It's clear that they could have the capability; it's not obvious from the article what they're doing with it.

I'm the "Senior Network Engineer" and I have full administrative control over every device (of ours) in our network. Thus, I have the capability to monitor any traffic passing through any point of our network. That doesn't mean I do, indiscriminately, 24/7.

Perhaps it's because of my role but I interpreted the statement, "We monitor all the connections", a little differently than you, I think. Replace "connections" with "links" and, like the man in the article, I also monitor all the connections. When a problem occurs, I want to know about it before we hear about it from our customers.

I think that by "connections", you assumed, for example, TCP connections. Unless they have a bunch of extra money laying around -- and it didn't sound like they do -- I'd be willing to bet they are NOT recording every connection that passes through their network.

Not every ISP is evil and trying to track everything you do -- and the small community-based ones like this are probably the last ones you should worry about -- and you shouldn't automatically assume that they are.

The quotes in the article seem ambiguous. Monitoring all the links, and monitoring aggregate bandwidth usage, seems perfectly fine.

I wouldn't expect them to keep logs, but I also wouldn't find it surprising if, when they observed a problem, they ran tcpdump on the gateway.

> the small community-based ones like this are probably the last ones you should worry about

I certainly wouldn't expect "evil" from people who have gone to these lengths to do something awesome. But I would expect a well-paved road of good intentions. A small community-based ISP like this would 1) not have an army of people telling them what will get them in legal trouble or create a PR incident, and 2) have the kind of "play nice" mindset that makes them not automatically think to treat customer log data and other personal information as radioactive. Unless the designers and maintainers of the network are specifically privacy-minded folks, the thought might not even have occurred to them.

In addition to the Big Brother stuff, I wonder why they even care how much bandwidth their users use.

As long as the links aren't congested, why care at all? Bandwidth is cheap and it's not like they are charged by the GB in any case.

Their upstream either charges a flat fee or by the 95th percentile. The Netflix example is really wierd, as usage is at it's lowest during the night.

Probably because their wireless equipment is the bottleneck and they are utilizing a mesh network topography which means bottlenecks impact other customers. The article stated that their backhaul uplink was on 70Mbps for the whole island.

Seeing as they're using Ubiquiti equipment, I'm not sure why they are not using of their Gbps setups for the backhaul at least.

Probably because they can't afford the backhaul. They pay $900 for 100M.

According to the network diagram in the article they do not appear to run a mesh topology. They run a hub and spoke topology. Each repeater has multiple radios.

Even so, they have serious problems if they can't sustain their 30 Mbps peak usage with their wireless equipment.

> I'm not sure why they are not using of their Gbps setups for the backhaul at least.

Considering they received a loan to purchase the necessary equipment to start this operation up, I'm guessing it's because they don't have a need for it -- yet, at least.

Bandwidth isn't cheap. But they should impose a traffic shaping policy to bring their users' transit levels in line with the fees they pay, rather than poking around into who is doing what on the network.

Disclosure: I built and run a small wireless ISP.

> Bandwidth isn't cheap. But they should impose a traffic shaping policy to bring their users' transit levels in line with the fees they pay, rather than poking around into who is doing what on the network.

That's the curious thing. The users are paying more than enough to cover all costs, so there is no need for all this tomfoolery. I guess they have either have mentality or technical issues.

Sure, bandwidth is cheap -- when you're buying tens of gigabits at a well-connected facility. The further from one of those you go, the more the price tends to go up. They're buying, tops, 100 Mbps, and having it sent to them via microwaves from a tower on the mainland.

> Sure, bandwidth is cheap

Indeed it is, especially when you consider their fees and costs. Based on the information in the article their monthly free cash flow is $2150, so they certainly can afford their current bandwidth usage be it at $9 or $30 per Mbps without playing silly games and spying on their users.

Why would you assume that they're spying on their users?

Because they implied so in the article, accurate as that may be or not.

A bit of an inconsistency in the article, perhaps?


> The monthly fees also cover the $900 a month DBIUA pays StarTouch for bandwidth.


> The StarTouch link uses burstable billing, with prices going up the more they use.

I don't think I've ever had any links that were, basically, a flat rate for n Mbps with the ability to burst over that for an additional amount, although that certainly doesn't mean they don't exist.

> "The Netflix example is really wierd, as usage is at it's lowest during the night."

Netflix usage or this particular network's usage? How can you know?

Burstable billing is where they use the 95th percentile as the basis for the bandwidth you used, so the top 5% are ignored and form the "burst".

I guess $900 is just the average bill for their usage right now.

Right, I understand how it works. I would have expected, in this particular case, though, that they have a fixed amount of bandwidth available at a fixed dollar amount per month (just for the consistent/predictable monthly expenditure). It may very well be that burstable billing was their best option, though.

> at peak usage times, total bandwidth usage across the entire network is 30Mbps or so.

The microwave backhaul is connected to a 100 Mbps port which would come out to $9/Mbit/month. From what I've seen, that's higher than what you'll pay if you're connecting to an upstream in a well-connected facility. If we then assume 95% billing and peak usage at 30 Mbps, that's an obscene per-megabit rate.

It's quite possible that you're right, of course. I'm making some assumptions based on my experiences which very likely are different from those on a small island in the PNW.

> I would have expected, in this particular case, though, that they have a fixed amount of bandwidth available at a fixed dollar amount per month (just for the consistent/predictable monthly expenditure).

Wholesale bandwidth is sold with a commit and burst. The commit is your fixed monthly charge (and bandwidth) you pay for each month regardless of how much you use. Overages, calculated with the 95th percentile formula, are paid at the burst rate.

> The microwave backhaul is connected to a 100 Mbps port which would come out to $9/Mbit/month. From what I've seen, that's higher than what you'll pay if you're connecting to an upstream in a well-connected facility. If we then assume 95% billing and peak usage at 30 Mbps, that's an obscene per-megabit rate.

Well, yes. I remember Hurricane Electric running a promotion for 1 Gbps of IP transit for $360 a month, but then again they are not present at a HE PoP.

As to pricing, it sounds like run of the mill rural charges. Also we don't know their commit, so can't really calculate with Mbps rate. Perhaps they just decided to go with a zero commit and $30 burst rate instead of paying a flat fee of $1500 for the 100M port. Would make sense at their very low usage rates.

>> "The Netflix example is really wierd, as usage is at it's lowest during the night." > Netflix usage or this particular network's usage? How can you know?

Both. Because that's how ISP bandwidth usage universally is. Peak usage starts after people get home and peak usage ends as they go to sleep. Few stay up during the night and even fewer schedule nightly downloads.

In many places in the country, especially in rural areas, backhaul bandwidth is actually both expensive and metered. "Bandwidth is cheap" is a rule of thumb that really only applies to places close to the major fiber routes.

This particular location is quite well connected, despite being an island. They have a local fiber co-op, a subsea fiber cable and line of sight to several mainland backhaul sites.

Are you familiar with physical internet infrastructure pricing, or are you just assuming? I don't think access is priced as bandwidth. I used to work at a small business isp and that was not how physical access was priced, particularly over relatively bandwidth limited physical connections. Though perhaps you work in the industry?

> Are you familiar with physical internet infrastructure pricing, or are you just assuming?

Yes, I am familiar and no, I'm not just assuming. I actually RTFA and I checked their upstream providers service offering terms.

and that upstream provider doesn't share pricing, plus if they are the only alternative to an isolate community, it's a sellers market

But the article shares pricing information and as I wrote in a sibling comment, the location is rather well connected for being an island.

It's going to differ around the world but in Australia for business connections I'm used to paying a separate fee for the physical connection (which will have a maximum throughput) and a separate fee for the bandwidth allocation and a third for the monthly data cap.

Australia is a bit of an outlier, mainly due to Telstra and the Gang of Four. I believe the polite expression for their actions is "maximizing shareholder value".

OpenBSD's pf does just that.

You get a guaranteed minimum but if someone isn't using theirs, you can have it.

And you can set all sorts of hierarchies if you felt like it. Putting users in say the SSH group and SSH getting 10% of all bandwidth.

Netflix has no auto play that I'm aware of. When the movie is finished, it stops. Good point about neighborhood watchdogs. I don't want my neighbors knowing what I do online.

Movies just stop, but TV shows will flow through to the next episode within a season on many Netflix platforms (including Chromecast).

It's really never made sense to me. Sure, play two in a row, but why would Netflix keep playing episodes any incur the royalty costs when people fall asleep watching a show, don't turn off the service when they turn off the TV, just plain walk out with the show playing, kids shows keep playing with no one watching?

I assume they incur royalty costs, or is there some incentive for them to somehow boost the number of hours of content played? Is this one of those cases where they are trying to inflate their numbers to look good to investors?

It will play 2-3 in a row without any user interaction, but if you haven't taken any action in a while it'll stop and give a modal "Continue Playing?" message

It does, at least on TV series it autoplays the next episode. Definitely does that on the Apple TV.

After a few (3?) episodes, Apple TV pauses the stream and asks if you are still watching.

My Roku does this after three episodes as well, although that adds up to quite a few megabits over a three-hour period.

Yeah, I thought this was really weird, too. I'm actually in the process of building out a small WISP outside of Georgetown, TX to serve an area that's right now /only/ being served by cellular, and I've gone with Ubiquiti gear. Not only does it allow traffic shaping but the practice is encouraged. There shouldn't be any need in a PTMP Ubiquiti setup to call people and tell them to lay off the Netflix.

Imagine "stop watching yourporn all night long mate" "we're looking at your traffic"

yeah.. no thanks.

> This is the kind of statement that makes me long for a faceless, impersonal ISP.

Slightly OT, but this statement reminded me of a question I've wondered about before and researched without finding a definitive answer:

Your average ISP probably has a lot of data which, in the wrong hands, would be excellent blackmail material. What regulations (if any) prevent the sale of that data to whoever wants it?

If you are in the EU there are pretty extensive data privacy rules. If you are in the US, it is a bit more complicated. It would mostly depend on what the US ISP claimed it was doing. If it did something beyond what it claimed it was doing, it could get in trouble. However, most ISPs are pretty broad with their claimed rights. Still a big problem with the bad ISP idea is that it would be hard to exclude children, and that could cause issues.

I think there'd be more of a problem if rogue employees get the data.

I think that you forget that most of the people on Orcas are retired, and don't have the same fears that we have regarding snooping of data. My parents retired to Orcas over a decade ago, and all of their neighbors are 60+. Also, their internet habits are very different than my household - they check emails once per day, and that is really it.

> the idea of a neighbor trying to "figure out" usage or "talk to" customers about it is horrifying

I sympathize (and would therefore avoid this size community in the first place, I think) but I'm curious: what would you have the customer-neighbors do, if they're being impacted by one person's selfishness?

> I sympathize (and would therefore avoid this size community in the first place, I think) but I'm curious: what would you have the customer-neighbors do

Throttle traffic, such that if there's more traffic than the total upstream bandwidth, everyone active at any given time gets a proportional share of the bandwidth.

And if someone's constantly at their limit?

To take an example from another domain, much of California's Central Valley municipal water infrastructure is unmetered. A consequence of this is that problems with water infrastructure go unaddressed. In most of the cases I'm aware in which meters have been installed, even where there's no actual usage service billing, a near-immediate result is to find many previously undiscovered leaks, mains breaks, or service pipe breaks. While I'm not an uncritical fan of "you cannot manage what you don't measure", there are times when having some sanity checks on usage help.

In the case of online usage, it may be that you're a spam hub. Or that, as at one hosting provider I worked discovered, the interaction of a client's autoplay audio and a particular version of MSIE meant that we were continuously streaming what should have been a 5 second audio clip. We overshot out 95%ile bandwidth cap that month.

> And if someone's constantly at their limit?

They still won't do any harm to anyone else. (Or if they do, then either the traffic shaping or the billing is broken.) Still not a good reason to track where the traffic is going.

However, it would make sense to give everyone fully-automated metrics on their bill, letting them know their total transfer, along the same lines as what the upstream providers provide. Make it clear that it's perfectly acceptable to use what they pay for, and that by design there's no information about the type or destination of traffic, but inform them that if the usage looks unexpectedly high compared to their own known usage (provide equivalencies in "hours of HD video" or similar so they have a baseline), they may want to investigate for themselves.

In the video, there's a shot from what looks like Cacti, an SNMP monitoring program typically used to monitor network utilization on routers (among other things). That's probably what they mean when they mention "connections".

Set guaranteed minimums for all users and only allow bursting for power users when there is spare traffic.

I'd rather have a mediocre d/l speed on a private connection than have my neighbors asking what I'm torrenting.

It makes me wish that Netflix, et al. provided a bit more in the way of configuration/control options. In this case, the ability to turn auto-play off. One checkbox, with a bit of code behind it.

Realize that not everyone is in the same boat, connection-wise. And with data caps continuing to loom, some would like a simple setting to help guard against accidentally running over.

Also, believe it or not, some of us like the credits. And with outfits like Marvel stuffing teasers and jokes in after they've run, that interest may be increasing.

Desktop: netflix.com | Your Account | Playback Settings | Uncheck "Play next episode automatically"

Mobile: Netflix App | Hamburger | Your Account | Playback Settings | Uncheck "Play next episode automatically"

Hmm. Wonder how I missed these. I recall going looking for such a setting, at least once. Thanks!

P.S. Found the described setting in the web client. Those settings pages look a bit different than I recall. I think they have added options to them since I last looked (months ago). Or, I am simply clueless -- wouldn't be the first time.

Thanks, again.

You're welcome :)

I looked for this on my Nexus 7 2013 running what I imagine is the current version of the Android Netflix app (no reason on my part that it wouldn't be), and I didn't find it.

Under the hamburger menu, I can only choose which user profile and what category of content I want to view. There is no "Account Settings".

The "three dot" menu does have a "Settings" item, but it only offers a very limited set of settings, not including the one you mention.

I did find a setting on the Netflix web page interface, and I changed that. But it appears to have had no effect on the behavior I experience under the same account and profile on the Nexus tablet.

(The web page Settings page offers many more choices than the Android app Settings page.)

Thanks, I never thought to log in to the website and look for a setting like this. We -- well, the girlfriend -- only use Netflix on a Roku and I wasn't able to find any settings for it there.

To be fair, "Netflix played all night" is not actually a thing that happens, as far as I can tell. When autoplaying, they pause every few episodes in case you're not watching. I'm sure they don't want to pay bandwidth/licensing for shows playing to an empty room.

I may be wrong on this, but I believe they are legally required to stop auto-playing every so often, so that it still counts as on-demand streaming and not simply a TV channel.

In fact, that's probably not correct, not sure where I picked that up.

Netflix has to pay for every "view" to whomever they're licensing content from, just as Spotify has to pay for every listen (although, the rules are much different for internet "radio" than they are for video streaming agreements Netflix has).

All services check if you're still around so they're not spending licensing money they don't have to.

I haven't allowed myself to run into it, but I seem to recall having read that it is after circa 3 episodes.

Nonetheless, if you are in the habit of dosing off, many nights, that can still result in a lot of "wasted" bandwidth, particularly with some of the more miserly caps being described in the (U.S.) press as having been imposed or "coming soon."

They provide that option. There's a checkbox under profile Your Account > My Profile > Playback Settings.

You can absolutely shut off auto-play from the playback settings screen in the netflix website. one checkbox changes it for all your devices though.

Not only that these atheist socialists are stealing revenue from God fearing business people.


FYI: most businesses in the tourist areas of Bali, Indonesia rely on wireless Internet by a small comapny owned by the Mega Internet cafe.

The national telco, Telkom, offers Speedy DSL, but it's not speedy or reliable, and Biznet offers FTTH only to dense commercial areas. (which is still more than Silicon Valley has!)

I had a tour of Mega a decade ago, and they have a tower pointed at Singapore. At the time they were using Linux LEAF distro for mgmt. and Soekris boards, but likely that has changed. Torrential rains do affect reception, but that also affects the Speedy landlines.

Also, outside North America, GSM rules. Which means you can go to the Borneo jungle and get 5 bars, which often doesn't happen in say, Sunnyvale. And 5 bars means adequate EGPRS data communications - everywhere.

I guess the residents of the San Juan Islands must be quite a different demographic from the Canadian islands just across the border. This would never happen on one of the Canadian Gulf Islands, because too many residents are worried about the "wireless radiation" causing headaches or cancer. (This is also where 1 in 6 residents refused a smart power meter [1], mostly out of supposed safety concerns.)

[1] "Gulf Islanders 15 times more likely to oppose BC Hydro smart meters than Vancouverites" http://www.vancouversun.com/technology/Gulf+Islanders+times+...

Regardless of unfounded concerns, I've refused a smart meter where I live for two reasons - one, they're extremely expensive, there's a large yearly fee equivalent to 20% of my power costs, even though they don't have to send a tech out to read the meter. Two, I don't want to share data about my power usage with my utility. I'm happy to have the data, and I do log power usage locally myself to a very high precision, but I don't trust the utility to keep that data safe - my power usage data is a VERY good predictor of whether I'm home or not, which is data very useful to burglars.

Beautiful place, I was there last summer. AFAIK suspicions about wireless radiation having adverse health effects have not been substantiated.

The subject comes up in places distant from BC. Considering the upcoming election year, here in Portland we usually have a few "fringe" candidates for local office asserting WiFi in public schools is "damaging" children, causing them to misbehave, etc.

Obviously most of the population is unconcerned about such risk. I suppose if it turned out that wireless signals really do cause some obscure deleterious effects, the "tinfoil hat" crowd will say they told us so. I'd go with the odds that's even less likely than winning the Megabucks lottery 3 times in a row.

Yeah, the only "proof" that people tend to cite is that the WHO has classified EM radiation in a class that basically means "we have no idea if it causes cancer or not, but it might", along with just about everything [1].

I lived in Victoria for a few years. My landlords there had refused a smart meter - yet they had WiFi, and both had cell phones.

[1] https://en.wikipedia.org/wiki/List_of_IARC_Group_2B_carcinog...

Here is one equipment from Ubiquity that you can use to built speedy point to point connection. 1.2Gbps on 100km link.


I'm guessing you haven't actually used one of these units? You'll be lucky to get half of either of those values.

FWIW, note that, unlike the majority of the network equipment industry, $UBNT's throughput numbers are aggregate values (i.e. equal to transmit + receive). 1.2 Gbps in $UBNT-speak, for example, means 600 Mbps transmit and 600 Mbps receive -- whereas an 1 Gbps NIC can transmit 1 Gbps and receive 1 Gbps simultaneously (yet we don't call them 2 Gbps NICs).

Up to 1.2gbps and up to a 100km link. I don't think you can get both.

How do you aim it at the other end? It's 100km away?

Same way you aim a Satellite dish to a satellite that's at least 35,000 km away.


For whoever downvoted the GPS answer, RTM. Please do, it's highly informative and very cool technical and educational material. If you refuse, please insert the F into the appropriate abbreviation and start this paragraph from the beginning.

Please make sure your loops converge. I read and modified this paragraph 3,984 times and then crashed.

Not sufficient. Civilian NavStar/ GPS has a 95th percentile accuracy of 7.5 metres and a best-case of around 2 metres.

That's more than enough to miss by several dozen dish diameters at 100km.

Main alignment with GPS and then finetune to get maximum performance. At 100km the main signal lobe is quite large, so being off by a few dozen dish diameters isn't large enough of an error to not to be able to get started.

at 100km it doesn't matter. Even with a beam divergences of 0.5 degrees 7 meters is literally a piss in the ocean:

Look at the polar pattern here: http://dl.ubnt.com/datasheets/airmaxyagi/airMAX_900MHz_YAGI_...

more importantly you have to fine tune the alignment anyway.

Thanks. I always assumed one could only get such distance with WIMAX

Or with a number of proprietary technologies(of which the airFibre is just one of many) - backhaul using point to point radios is used a lot in telcos and there's numerous (expensive) products available.

In practice it's more like 10 to 20 km for max throughput.

This is pretty awesome, but I wonder if it could have been done had there not been a network guy in the group? I've done a lot of network code/config over the years, and I wouldn't say it was particularly easy. The guy seems to say it was easy.

Unbelievably fiddly is more like it. Just the nature of it makes it annoying: before you have a connection through, you have a chain of links that doesn't work. You have to fix every single link before Eureka.

> but I wonder if it could have been done had there not been a network guy in the group?

More to the point, what happens if he moves away or is unable to support it.

I do wonder if there is a business model to implement and support this type of network for small towns.

> I do wonder if there is a business model to implement and support this type of network for small towns.

Well, you can outsource the NOC, but no amount of remote handwaving is going to help when you need hands on-site. In the end you get what you pay for.

>In the end you get what you pay for.

Well, they were paying for CenturyLink and got a 10 day outage.

In all fairness that outage was due to a subsea cable break and those require both specialized equipment and time to get the equipment on-site.

... and it's a really good example of getting what you pay for. I really do not think the islanders are paying Centurylink enough to be able to fix subsea cable faults with a 4 hour lead time.

That being said Centrylink's lack of redundancy and fallover links is another matter.

In my city, actually 2 different ISPs were started by a bunch of hackers trying to provider better internet for their own peers. It seems to be at least not impossible to do so.

Building and running an ISP is the easy part. Being allowed to do so (regulation/facilities access) and making money are the hard parts.

Well, at least in Germany that is a lot easier – big ISPs have to rent you their lines at just above cost if the lines are older than 2 years.

Which led to people often having dozens or more ISPs to choose from.

Good one. If you are lucky enough to be in a city. The majority of the cabling and infrasturcure was built by the Deutsche Post, when it was still government-owned. Now nearly all of that belongs to the Telekom, which have a massive tmonopoly, even though they have to rent their cabling too other companys for a low price. Problem is, there are very low infrastructure costs for everyone at the moment, leading to a very very fierce price battle at the internet market, which leaves no money for improvements of the infrastructure.

In short, in Germany many people skipped from Modem/ISDN to LTE because they had no other choice for 15 years than to stay with Dial up connections.

> The rural Orcas Island has a lot of hills and obstacles that could disrupt the wireless signals, and it would have been "prohibitively expensive" for DBIUA to install its own towers. As such, many of the radios had to be installed in trees.

Presumably there will be movement of antennae as trees grow. One could monitor the alignment by periodically checking the signal strength between links. I wonder how often things will need to be realigned because the "tower" on which the radios sit changes with time.

Depends on now close to the top of the tree the antennae is placed. Trees grow from the top, so anything below that won't move as the tree grows.

As the tree grows, the weight distribution changes, and things move. Trees grow from the leaves. Leaves are not just at "the top".

Trees also grow around the trunk and branches, thickening. Ever heard of counting the rings on a tree to tell its age?

even highly directional antennas arent, its usually >7 degrees

Once they pay off their loan, they could end up with the cheapest and faster internet services in the area.

The irony.

Well, a business would take the free cash flow and invest in expansion. Then again, a business would be paying the island's network operator and there would likely be no free cash flow.

I think what may happen is a big operator will invest a lot of money to make service there incredibly cheap and fast, killing the service.

As a lesson to all you out there who'd dare challenge a communication company.

This is really cool!! I always wish to setup such network amongst peers and get everybody connected with good b/w, especially in India. This kinda stuff can create marvels in emerging economies whose existing infrastructure is heavily loaded.

This is really cool. Honestly I'm surprised they still get 20Mbps after the data goes through multiple relays and a microwave connection. Are microwave links in widespread use for internet communication?

There's a whole group of companies called "Wireless ISPs" [1] that make extensive use of microwave products from companies like Ubiqti and Mimosa. Depending on the backhaul, you can get gigabit speeds over unlicensed microwave spectrum in many locations. There's a WISP in San Francisco called Monkey Brains that is a really good alternative to Comcast if you have the space for a microwave antenna on your house. [2]

[1] - https://en.wikipedia.org/wiki/Wireless_Internet_service_prov...

[2] - https://www.monkeybrains.net/wireless.html

In SF, I'd also recommend Webpass [1], which uses point-to-point microwave to get to apartment buildings and then piggybacks on phone data lines within the building. They're cheap (by American standards - $500/yr), fast (from 100/100 to 500/500 depending on the building, usually 5ms ping), and have great service (no installation fee, tech was on time to the minute and extremely knowledgeable).

[1] https://www.webpass.net

I'm also super happy with WebPass. My current building gets about 400 Mbps up/down.

Huh, I thought Webpass was about fiber to a hub for the building.

I actually use a Wireless ISP that uses a little satellite dish. I don't know how well it works or what kind it is, but I have mostly slow speeds, and upgrading is more expensive than it should be.

Just for reference, it uses Cambium Networks' PTP-100 by Motorola. I've actually only just discovered this today and am not entirely sure what all of it does.

Assuming it's PMP100 (PTP100 is essentially the same gear, except a pair of radios dedicated to each other instead of one AP and up to 254 clients).

PMP100 is capable of 14Mbps aggregate, often set to ~10Mbps download, 4 Mbps upload. It's roughly 10 years old at this point. The main benefit over WiFi, Ubiquiti, etc is it uses FSK modulation instead of OFDM so while it is slower it can go farther and tolerate more interference.

Cambium has a newer product, the PMP450, that is capable of quite a bit more throughput. Right now IIRC it does ~125Mbps per sector but using OFDM modulation so you need better signal to noise, usually achieved by limiting distance to clients.

The client radio on PMP100 costs ~$199. Client radios on PMP450 are license keyed by sustained capacity. A radio with 5Mbps sustained is ~$199, but unlocked ~$450 per.

Many Wireless ISPs operate on tight cashflow and have trouble self-funding that with normal residential pricing so they want customers to pay for gear up front and that tends to cost too much for consumers.

Yes - I was surprised when my Son found out that his apartment complex in Boulder Colorado linked to Denver via a microwave link of 30 miles or so to get on the Internet backbone. Fiber abounds around here, but It must have been cheaper than leasing a line. Performance wasn't that great which is how he found out about the link :-(

Is he in WilVil? That's the only apartment building I can think of that would be tall enough to conceivably get line of sight to Denver. Of course feel free not to answer if you feel weird about the question (ie. location of your child).

He doesn't live there anymore, but the complex was Two Nine North. I suspect that the transmitter was not on the complex itself, but instead was at some location the complex's ISP leased. He was having intermittent/nightly dropped packets, long ping times etc. He complained enough that they eventually passed him on to the technician that configured the microwave link to the Denver Tech Center area and at least to my Son, it appeared that it was set up very poorly.

I'm rather surprised by this. Internet connectivity in Boulder is really quite good[0], since the University drives a lot of traffic.

[0] Spent the last 11 years in Boulder.

His complex, Two Nine North, likely chose the lowest bidder

Don't confuse high bandwidth for low ping. There are many high bandwidth high latency services - like satellite uplinks.


I would like to be able to share my home network with my car within a 15+ km radius of my home... Maybe using a servo-controlled directional antenna on my car that would always point to my home using some sensors, or an unidirectional would be even better but I don't know if that is possible using open-spectrums.

Considering the investment you'll need to make in antenas, equipments and such, unless you live in a poor coverage area, you'll be better with a 4G.

That remains to be seen but I have a feeling that savings could be seen in the medium term, or maybe just a few years... 4G is too expensive and data caps are ridiculous.

Yes, and if you need connectivity between the two, setting up a VPN would be easier.

Hope it works out well for them. I had a horrid time with a local wireless provider in my area. It was based on the motorola canopy (now cambium networks) configuration. They had main antennas on every single water tower in probably a 20-30 mile radius. I live on the edge of the county and there is not much out here. The water tower antenna that I was pointed to was probably servicing a ton of people.

When I first got set up, I was told I would get 8mbs down and 2mbs up for $70 a month. It was unlimited, unmonitored, and a "straight static IP into the internet". For the first 3 months, it was great. I got 8mbs and pings were 80ms. However after those first 3 months things, started to slow down. My antenna was in front of a tree and since it was now spring, the tree was covered in thick leaves. I thought that was the reason for my slow down and I didn't complain. Once winter came again, and all the leaves fell off, there was no difference. My speed was 8mbs, then it was 7mbs, and now it was hanging around 6mbs and I could no longer use my VOIP office phone. I could hear fine, but no one would hear me. And also during the evenings, things were getting worse, with speeds dipping down in the the 4mbs range, even netflix was starting to slow way down. Another year goes by and I am looking at a max speed of 5mbs, but normally it dips down to 1mbs. Another year goes by, and it was even worse, but now pings are up to 100-200 range.

I then noticed after these three years, ATT u-verse DSL was available. I managed to get the business class 6mbs for $50 a month (but I get 7mbs down). I've never been so happy....

In hindsight, I probably should have complained to the customer support, but I really had no faith in them, it felt too much like an over subscription issue and I was afraid that their "fix" would be to just install a 50 foot pool in my yard to put their antenna on which I didn't want.

Other than that, I have always wanted to build my own ISP. But now I don't know if I would do it wireless or not (but that is pretty much the only way...)

CenturyLink has gone so far as to tell customers who cancel their DSL service that they will not be able to start it up again, Brems said.

That seems awfully abusive.

The costumer doesn't want them because the service constantly fails and they don't want the customer because he constantly complains. What's abusive about that? A perfectly healthy non-relation. The only abusive thing would be gov. getting in the way of it and forcing them to accept the customers or the costumers to accept the ISP.

CenturyLink sucks in general. Poor service and poorer customer service. That company is a joke. Seems their whole purpose is to drive people with wired phone service to other companies using wireless.

I don't think they're trying to be abusive on purpose. It seems like they're running out of capacity and desire to serve the island.

All the more reason to pull the rug out from under the legacy operators

Now that ARIN's run dry, it's gotten more difficult to start a proper ISP. Does their network support IPv6?

> Now that ARIN's run dry, it's gotten more difficult to start a proper ISP.

Only if you want static IPv4 addresses. And even then, they aren't impossible to come by.

Dynamic addressing only helps if the maximum number of simultaneous users is significantly less than the total number of subscribers. When you're connecting houses, that's generally not the case.

You don't need to hand out routable IPs at all. NAT should support roughly 64k simultaneous connections (not users) per IP.

The Web is not the Internet. If customers have no means to receive unsolicited packets, then you're not really an Internet service provider.

Is that guy using ubuntu on his laptop ?

It does look like a unity interface, so yes I think so. But why would this surprise you? The guy is a software engineer after all...

I'd expect Arch...

Why? Lots of us "techies" don't like Arch.

I wonder how they set this up w.r.t regulations. I was under the impression the process to setup an ISP involved a mountain of red tape. Did they get away with it because they are a non profit?

There's very little red tape involved in starting as ISP in the US. The red tape starts up to 6 months in when you hit the bi-yearly FCC Form 477 filing, though I'd say the majority of (w)ISPs are flying under the radar by simply never filing.

If you even think about offering voice services (even VoIP), then you get buried in paperwork.

Those laws vary by locality (e.g. it's typically a per state thing). I do believe Washington is pretty non-regulatory when it comes to ISPs.

I see the a headline “It wasn't that hard”, Then immediately below a picture with the caption "Sutton holds a drone he used to analyze potential radio location"...

They should register here: http://db.ffdn.org/ !

Is 3.65 GHz (mentioned in the article as a way to get away from crowded 2.4 GHz) allowed for unlicensed use in the US now?

looks good but i believe they can take more learnings from http://battlemesh.org/ for creating these kind of adhoc networks

Registration is open for Startup School 2019. Classes start July 22nd.

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