@rune @rysiek @Hyolobrika@fedi.club @sofia OTOH you could decentralise DNS somewhat by replacing ICANN with a consortium made up of the national registries, and require a number of root servers at each of them. At least as an outline of a solution, that is.

Follow

@rune @rysiek @Hyolobrika@fedi.club @sofia The point here is, it's theoretically very easy to decentralise DNS, because other than the root maintenance, it's already decentralised - via delegation.

@jens @rune @Hyolobrika @sofia and, spoiler alert, this is kinda sorta where we are now. Almost.

#ICANN was forced to create certain governance structures and procedures that make it possible to get involved and have a say in their decisions. Plus, with the global scrutiny on them for quite a while now, they are really doing their best to tread lightly.

It's far from perfect, but way better than some blockchain-based techbro controlled contraption that would have curious bugs in its smart contract.

@rysiek @jens @rune @Hyolobrika @sofia and even with ICANN in place there are still independent root servers that serve own TLDs 😱

All it takes is adding one line into your DNS resolver's config and magic happens.

Not to mention that national domains are also not simply ruled by ICANN, these organisations are able to make independent decisions. And thanks to "Too big to fail" a lot can't just shut down by ICANN.

@jens @rune @rysiek @Hyolobrika @sofia Actually, DNS is distributed, and far from being decentralized. A tree is, by essence, hierarchical, and hierarchy is, by essence, opposed to peer systems. One could decentralize one node of the DNS tree, and a branch could be composed of decentralized nodes, but each delegation acts as a bailliwick, both technically and organisationally.

@x_cli @rune @rysiek @Hyolobrika@fedi.club @sofia That's a lot to unpack.

Definitions of decentralised Vs distributed go back to Baran's 1964 paper ( rand.org/content/dam/rand/pubs ), where the main criterion he uses to distinguish distributed systems from decentralised ones is whether the destruction of a node or link affects the availability of nodes (in a nutshell).

Conceptually - in name structure and resolution order of full names - DNS..

@x_cli @rune @rysiek @Hyolobrika@fedi.club @sofia ... is not in the slightest distributed. What makes it distributed technically is that at every level in the name hierarchy it's possible to have redundancy, such that destruction of any individual name server does not affect resolution.

Organisationally, it's not distributed, either. Here, we don't even have this kind of redundancy. It's not as if - commonly speaking - any name...

@x_cli @rune @rysiek @Hyolobrika@fedi.club @sofia ... component is managed by multiple legal entities. So I don't really know where the claim comes from that it's distributed.

It's very much decentralised organisationally, though, in that no central entity controls the entire name assignment space.

TLDs at the root are a bit of an exception simply because they're a single root, because the names are hierarchical.

Are we talking about...

@jens @rune @rysiek @Hyolobrika @sofia I suppose this is indeed a terminology issue. DNS is not decentralized because there is a very clear "center": the root zone, and every domain name acts as the center of their own subdomains. However, callling the DNS centralized would be inaccurate, because TLDs and specifically ccTLDs are owned by various entites or countries, each having their own set of rules, legislations and technical infrastructures.

@jens @rune @rysiek @Hyolobrika @sofia
Now, is "distributed" the best adjective to describe that topology? I believe this is the case, but I can accept that one would disagree :) Clearly, DNS is a client/server model, and thus, one could argue that the client/server model is by essence not distributed. I would say that it is "distributed enough" :D

Sign in to participate in the conversation
Finkhäuser Social

A private instance for the Finkhäuser family.