Could be interesting to look at Patchwork/ScuttleButt for inspiration. It uses a 'gossip' protocol that primarily operates over LAN connections, and has a bunch of other cool stuff.

There's a rather large 'escape hatch' to the local nature of it all in the form of 'pubs', which are basically regular users that are internet-accessible.

But at least for the interface and some of the decisions they made, it might be worth investigating.

That's interesting—I'll definitely give Patchwork a try. I'm curious to see which 'Pubs' have been created in my city. That said, the focus seems pretty distinct from my own interest, which is basically to create a kind of forum where the top-level 'topic' is some real life location/object. Patchwork's interface does look really nice though—thanks for pointing it out.

Well, Patchwork is just one possible app on top of it. There's also a chess game and a soundcloud-ish app. So perhaps it would be possible to make what you have in mind on top of the protocol.

The main problem I have right now that keeps me from playing around with it is that I'm not currently much in the mood for Node.js development, and so far that's the only really solid/easy implementation of the whole thing.

I definitely agree with your thoughts on locality and anonymity! However, I'm also of the opinion that decentralization via offline device-to-device communication is a key component.

1) Barrier of entry. Having to invest in a hardware beacon to kickstart your community (as well as having to go to a physical location to access it) means that "hit and run" communities and users won’t be too much of a problem. You can see this working in communities like Metafilter and SomethingAwful which charge a nominal entry fee to keep low-effort users out. But notably, the beacon wouldn't be essential: once a community is formed and the peers all know about each other, they could connect directly without having to touch the beacon first. (For convenience, it might still be useful to keep around for the same reason people use Github.)

2) Persistence. Personally, I’m reluctant to seriously commit to any centralized social networks, and I think many others feel the same way. Startups rise and fall with the blink of an eye, and it’s disastrous when the community you’ve spent years cultivating suddenly vanishes due to corporate acquihires or pivots. If a community is based on peer-to-peer protocols, and if the data for that community is stored locally on each member's phones and computers, then the life of that community is completely decoupled from the whims of its creators and the fate of its tech stack.

3) Aligned incentives. Peer-to-peer communities are beholden to no organization other than themselves. They can determine their own fates. They have nothing to fear from shareholders or CEOs. Nobody can make them function differently without their consent. And vitally, by design, there can be no temptation to take the community online for convenience. Meeting up at a physical location is an inherent and inalterable part of the system. (Though of course, one peer could manually relay the state of the community to another peer through an online channel. But that wouldn't be the typical use case and would require manual work.)

I admit that perhaps it's more of an emotional issue than a practical one. Even though sites like Nextdoor (or the same AnonCon) are tied to physical locations, all the data is still stored on some central service and could easily disappear tomorrow. Whereas a mesh-network-style community would in fact be an artifact of the real world, with all the pros and cons that this entails. Different behaviors — ones that I can't predict — would surely emerge from these differences in implementation. (But as a starting point, a central-server-based, geo-pinned community service would be really awesome, too! I'm just not sure it would be possible to deal with location spoofing.)

I like your interface idea! I can imagine "terminals" appearing in the list much like Wi-Fi hotspots or Bluetooth devices. If they were indeed paired to physical devices, perhaps some central site could help keep track of them (à la Geocaching).

I wonder if a Raspberry Pi could be seen and accessed by an iOS or Android device without an internet connection?