Standardize on a iroh connectivity URL

[dpc]

Sorry if I missed an existing discussion, but we are planning to encode Iroh connectivity as one of the connectivity options in our application, and since we're doing it, we would rather do it in a way that has a chance of being somewhat standard, and not just entirely ad-hoc.

So I wonder if anyone thought about it and if there's any precedence.

What we consider now is:

fedimint://<iroh-id/pk>.iroh/
ws://<iroh-id/pk>.iroh/some/path
http://<iroh-id/pk>iroh/

etc.

So the actual protocol is still indicated via scheme part of the protocol, and the iroh connectivity is indicated by the iroh TLD, with the domain being a (somehow) encoded public key/id. This is modeled after .onion services which seem to be in the layer/function architecturally.

Does this make sense?

The details to agree on are the encoding of the ID, and the exact TLD, as it would potentially conflict with standard DNS hierarchy.

[flub]

Hi, interesting consideration. So far I think our view has been that a URI scheme is out of scope for iroh. The only thing that you know about two random iroh nodes it that they can connect to each other on the QUIC level. But for an URI scheme you need application-specific semantics, like you need to know which protocol(s) the nodes will be able to speak over that QUIC connection. So URIs are really the application domain.

Personally I'd be somewhat careful about using domain names in URLs which you (or we) have no control over like your http and ws examples. Maybe this also includes the fedimint scheme, but there you'd control the scheme so it is not necessarily followed by a domain name.

What I have done in an http3 expriment before was use something like iroh+http but that was very ad-hoc and experimental. I'm not really sure what the RFCs say on URI schemes.

[dpc]

Agreed. Iroh is on a transport, not application level. iroh+<protocol>:// is something I've considered as well.

I've used <pk>.iroh domain because this mimics what Tor's Onion services are doing with .onion TLD. As far I can tell there is no iroh TLD: https://www.icann.org/resources/pages/tlds-2012-02-25-en , so it should not conflict with standard DNS hierarchy.