Follow

how could we eliminate the reliance on dns for identity

i'm thinking dht-based network but i don't have 100% conceptual understanding of that field

each user gets a hash; each hash in the table would point to that user's current address/location?

and how could this be done as an extension of activitypub

i'm thinking maybe a "urn" field like how "url" is distinct from "id"? either that or making activitypub not require http somehow

i guess the other issue is how to translate from urn to location, or how to find an object purely by urn. you would need a resolver of some kind

@trwnh A few thoughts here, on this whole thread:

1. .onion addresses are DNS without the DNS part, and it's one way to get away from the DNS issue in the short term.

2. DIDs are another way to consider identity online.

3. Removing HTTP would be doable but HTTP isn't so terrible either.

4. I've been thinking about DHTs in relation to DataShards. Happy to chat about it.

@emacsen hmm.

1. not so sure that works? it requires being on tor, right? so it wouldn't work over clearnet

3. i dont think http is bad either, but its reliance on authority/hostname presents issues as above

i'm really interested in how to resolve a urn / how urn namespaces are assigned, mostly... how might a global urn system work?

@trwnh I'd love to understand your thinking more. Your original message was about eliminating the need for DNS on identity, and then you talked about DHT networks.

Let's take a step back and ask a more fundamental question- What is the goal?

Is this just about DNS? Is it about privacy? Is there something else?

And what in the existing AP alternatives do you dislike, and what are the properties you feel are indispensible?

@emacsen just spitballing here. my chain of thought was kinda like this:
- ugh, "id" being https kinda sucks
- but it needs to be resolvable
- how do you map names and locations together
- actually wait maybe you could just resolve names directly somehow?

not sure there's a concrete goal here except to allow for referring to things without knowing their location

@trwnh If what you care about is exclusively the goal of not referring to things by their location, I go back to suggesting TOR Onion Services, but also Secure Scuttlebutt is worth investigating (we did a show about on @librelounge ) and Briar is also worth checking out I think.

@emacsen I mean how to do so as an ActivityPub extension or similar, over clearnet.

@trwnh pretty sure tox.chat/ does DHT on hashes of public keys.

It doesn't really solve the naming problem because hashes aren't names, though. With Tox display names are filled-by, and marked-as suggestion and you can accept them, if i remember correctly.

That doesn't sound workable for mastodon, that said, you could just accept a hash-as-domain-name as adequate?

@jasper public keys can be cycled out. that shouldn't change your fundamental identifier.

i'm just thinking of how we have contact books for phone and mail, but not for social networks. but you say "i'm [username] on [network/domain]" vs "my phone number is [number/identifier]". centralized services do their own mapping, but we need decentralized mapping too. but what would be necessary to make it work, with authenticity?

@trwnh note that the DHT is not really that important for naming here, it just finds out how to connect to a particular public key..

I think naming things inherently requires consensus of some sort, and we pretty much have authority or blockchains for it right now?

Though suppose authority could be multiple instances and it's true/another voting instance is added if they all sign off on it.

Also blockchains don't need to have PoW/PoS rules for determining who can make the next block.

@jasper we don't need consensus if the mapping is coming from someone you trust. "oh hey, this is x's phone number." or on a larger scale, the white pages.

@trwnh I seem to recall that Hubzilla at one point claimed to be DNS-agnostic when using Zot? I'm not sure what that entails, but @macgirvin might be able to explain it.

@sean @macgirvin well zot relies on identifiers and it uses id for all references inside the zot network. but i'm wondering how to translate references *between* networks. i'm not too familiar with DID but probably something in that direction -- did:activitypub:id or whatever, which can coexist with did:zot:zot_uid, and then you add mappings to your address book.

@sean @macgirvin the main issue is with HTTP and authority. what do you do when someone doesn't have a DNS domain name they own? you can fall back to authority by IP but that's even more fragile. with activitystreams/pub relying on HTTP URI it's hard to separate authority in a way that works for everyone.

Sign in to participate in the conversation
Mastodon

Server run by the main developers of the project 🐘 It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!