Comment by contingencies | Hacker News Reader

Comment by contingencies | Hacker News Reader

Parent Full thread

Mesh is a mess because the government, mobile device manufacturers and cellular carriers all want you on a centralized cellular system. Money is made selling cellular plans, cellular data, cellular carrier based device distribution (especially in the US, where most devices are carrier-acquired IIRC). Google in particular don't want to encourage non-centralized models, because their whole Android-shall-prevent-Apple-from-taking-the-mobicake play ("commoditize your complement"[0]) was to preserve their cloud kingdom. They're using lawyers and doublespeak to lock down what they've got, for example through ensuring no app can be distributed on Google Play that supports non Google Play store based payments (eg. carrier payments), and no app can be sideloaded without great effort (and there you were thinking it was your device!) At certain levels of their organization they are actively refusing to make mesh work in new Android versions as a first-line use case in Android, even though the hardware has supported it for eons (since ~2011[1]), as well as fiddling with the Android security model enough to make it a pain to use even if it theoretically works. This seems to be a dual layer strategy to frustrate any non-edgecase popularity of mesh. They do claim to support it in home networking products but probably only because they want to take over, monitor and surveil the rest of your house to understand more about your family's behavior, presence and so on. People even leave the bloody microphones on all the time, knowing it's spying on them. It's insane. This is why we need open devices.

To be fair, the security model for mesh is hard. Even FirefoxOS couldn't get mesh working properly. I think someone needs to just bring out an open device with mesh as the default use case. If anyone wants to pile on and do it, I'm down to contribute.

[0] https://gwern.net/complement [1] https://en.wikipedia.org/wiki/IEEE_802.11s

Mesh is a mess because meshes do not, in fact, work very well.

You set up a mesh network to provide community internet access to an underserved community? 99.9% of traffic will be going to or from the node that connects to the internet at large.

That means that although notionally adding a node to the network adds more bandwidth to the network, in practice that's only true if the new node happens to add extra bandwidth at a bottleneck.

And if everyone aims their directional antennas at the egress node, to avoid this? You've just reinvented the centralised network.

I don't understand how anyone would expect to gain more bandwidth from a mesh. What you gain is coverage, as in distance from uplink.

And at least with earlier meshes, you decrease bandwidth even if you use only one device, because now in laptop<--wifi-->AP<--mesh-->router both links compete for the same airtime.

> I don't understand how anyone would expect to gain more bandwidth from a mesh.

Well, if you had an hourglass-shaped mesh network which was congested at the narrow point of the hourglass, and you recruited an extra node that gave you a second link across the narrow point, by adding the node you've doubled the bandwidth between the bulbs of the hourglass - and improved the redundancy.

The problem is a lot of people imagine this will happen every time they add a node, making the whole thing infinitely scalable. Whereas the reality is any time you add a node that isn't crossing the narrow point of the hourglass, the congestion gets worse.

You're talking about a world of P2P and local use.

The reality is that practically everyone is using "wifi mesh technology" as range extender for their own singular internet connection, and the internet connection is the bottleneck for practically everything.

Yeah, with current consumption being "everyone wants a global DNS and global network access" it's semi-pointless. But that's just a chicken and egg problem. All of Wikipedia, libgen, thingiverse, open source packages, etc. can be locally cached somewhat trivially. Most oldschool games will run P2P fine over arbitrary networks. That's quite powerful indeed. Now we just need some localtalk apps (reinvent 'talk' and BBSs and get people off Failbook) and there will be users. Also, piracy can be normalized in a social way instead of the current situation so people can actually, you know, discuss and appreciate art. A killer app may be sharing your CCTV with people, so the community as a whole can more readily resist bad actors. Perhaps a starter node could be a bit of media to pick up wirelessly on your way to your commute, so you have a kind of protected local distribution that's censorship free and surveillance costs substantially more than with centralized networks. Next, bring back Fidonet! Haha. Maybe one too many red wines...

If only the users were better and didn't want to stream video, or talk on Facebook, or video call each other, I could run a brilliant service!

Just need a decentralized system where people could be rewarded financially for providing network access.

Imagine a plug and play system where people with a satellite link could sell network access to their neighbors without a headache. Clients would just need to specify how much they are willing to pay for GB and nodes how much they want to get paid in their routers.

There are many great things that could be done but AFAIK nobody is investing in making that a reality because governments and those that benefiting from the established status quo would definitely get in the way. So Starlink is the way as there's not much that governments can do to get in the way of thousands of satellites controlled by company outside their territory.

It seems to me an approach would be to purposely deoptimize such a P2P network for these applications. Optimise for time-insensitive best effort delivery like the internet of old. Still plenty of applications can work to that model and it is “good enough” for doing business.

Keep the bandwidth heavy applications on different channels with a different QoS guarantees and cost structure.

I put “mesh” in quotes for a reason. I mean the wireless backhaul ability that basically every major WiFi AP manufacturer has, not a bunch of unaffiliated users forming an ad-hoc, dynamic mesh.

Don't assume a conspiracy where the underlying mathematics are a sufficient explanation! Centralized systems use RF bandwidth MUCH more efficiently.

For long-distance, high-power communication, sure! For local communications, one big counterexample is Wi-Fi – it's centralized on a topological level, but decentralized in terms of spectrum assignment.

And even for longer range, that reasoning can't explain the absence of anything at all from Google or Apple in off-the-grid situations. Why can't I text (iMessage, WhatsApp etc.) a friend sitting a couple of rows away in an airplane without Wi-Fi, when I can airdrop them gigabytes of data just fine?

Yes quite right, I don't doubt it's the organizational behavioral paradigm at play, not necessarily an organized conspiracy. A key quality of an established bureaucracy is that it can arrange, design, deliver and entrench bad outcomes without being aware, certainly without actively conspiring. However, the personas that float to the top (or even the middle) of these things are pretty off-putting, and they're also aware regulators will be on them if they get too explicit, so I wouldn't like to place sustained, conscious, premediated decision making out of the realm of possibility.

Reply on news.ycombinator.com