Re: A native internet protocol for social media

>> Attacking DNS is a perpetual red herring here, and enda up with Zooko's
triangle failure modes a bit like that URL. You have either alternative
resolver roots that are more proprietary than DNS, content addresses that
are fugly, or more fragile key management models with worse failure modes.
DNS is a success.

> It is possible to make content addressable identifiers on the web more
user-friendly by embedding them in hyperlinks, which can be more easily
understood and navigated by end users.

It's also possible to use DNS as a petname system for the "real"
identifier. Who cares if the "real" identifier is "fugly"? It's not for
you, it's for the machines.

One historical thing I wish we drew more inspiration from is the XRI
concept of "i-numbers" vs "i-names", as well as subjects and aliases at the
DNS resolver level. In practice today we have Webfinger which can
theoretically fulfill the same role -- ask it to resolve a DNS-based
resource URI and get the not-necessarily-DNS-based subject URI from the
resource descriptor. For example: if you resolve an https: URI that is an
alias for some other identifier with a non-DNS root-of-authority (URN, DID,
etc.), then the https: URI never has to be used directly by anyone except
the user and the user-agent.

Put another way: DNS is a "success", but there is still a "failure" in that
it is reassignable and open to insolvency. I would not use DNS/http:/https:
as the canonical or authoritative identifier for anything I wanted to
outlive a yearly DNS lease. Sure, you can refer to me as ""... but
only for now. We need another layer of indirection to allow that domain
name to change.

Received on Wednesday, 12 April 2023 13:48:36 UTC