3 ms·
No, P2P is impractical and it inevitably leaks into the UX, making it significantly more cumbersome. With a P2P system, you can't just send someone a link to so
by grishka 5y ago
No, P2P is impractical and it inevitably leaks into the UX, making it significantly more cumbersome. With a P2P system, you can't just send someone a link to something you found — you either expect them to have a client for the service, or use a web proxy.
> Sign and publish with a pubkey
No, absolutely not. Public keys make terrible identifiers. You leak it and anyone can impresonate you, with you having no recourse. You lose it and you have to start a new life basically.
Account recovery and the ability to revoke access are not "nice to have" things — they're hard requirements for any identity system that is to be used in the real world by real, including non-tech, people. Trust me, I worked at the largest Russian social media company, they have a dedicated account recovery department for a reason.
- Ambolia 5y agoCouldn't this be solved by chains of trust and "blacklisting" accounts? This is pretty much how it already works in the most annonymous corners of the internet, and those are the best corners.
- echelon 5y ago> No, absolutely not. Public keys make terrible identifiers. You leak it and anyone can impresonate you, with you having no recourse. You lose it and you have to start a new life basically. Most average people can't even get the platforms to deal with ATOs. I'm not into crypto, but it's also accepted in that community that you must safeguard your keys. But beyond the status quo, I can easily imagine dozens of mitigations and protections and additional proofs: revokable cert authorities and peers, multiple signatures, portable / updatable keys, etc. etc.
- grishka 5y ago> I'm not into crypto, but it's also accepted in that community that you must safeguard your keys. Of course it is accepted, but so it is with passwords. And people are terrible enough with passwords, yet here you are suggesting that they use something that can't even be (easily) memorized, can't be invalidated, and has to be stored reliably yet secretly, as "the" identifier. You're putting too much trust into your average user. None of these cryptography-based identity schemes can possibly work in the real world, period.
- Ambolia 5y agoYou could still have server-owned accounts with that setup. It just wouldn't be the only option.
- grishka 5y agoYou could still have public-key-identified accounts with ActivityPub, too. ActivityPub does, in fact, use public-key cryptography for authentication, and each actor must have a key pair. It's just that it's not its identity — the identifier is a URL where the actor JSON object can be found.
- easrng 5y agoIn practice the keys are managed by the servers though.
- lumost 5y agoThe only time these systems will work is when the perceived risk of snooping is greater than the value of account recovery. There is certainly a niche for this, but it probably isn't very broad. The majority of internet users seem uncomfortable with surveillance, but accepting of it. I suspect the pendulum will swing if users no longer feel they have anyway to opt-out of surveillance. e.g. private browsing and other systems are recognized as ineffective, or sensitive data such as chat messages become broadly leaked and accessible to the typical user.
- DarylZero 5y ago> With a P2P system, you can't just send someone a link to something you found — you either expect them to have a client for the service, or use a web proxy That's not a property of P2P, it's a property of anything that's not built into browsers. That is, either the protocol is built into browsers and the browser is the client, or else it is not built into browsers and something else is the client. Modern browsers already support P2P through RTCPeerConnection. In the future there may be even more general P2P capabilities built into browsers. What you're talking about is at most a limitation of current browsers, not of P2P.