Hacker News new | past | comments | ask | show | jobs | submit login

I love the discussion & effort.

On the other hand, relying on proof of work & creating an economy just to have distributed naming, to me, seems unnecessary & limiting. Users ought to be able to generate & use names heavily if they so desire, without having to do meaningless work to do so.

Systems like wireguard and HIP aren't name systems but they are connective systems, built around cryptographic identity. They assume a system of plenty, where folks & systems can be generating & using new identities regularly. To me, i'd much rather create a name system build around identity & assumptions of plenty, than a more name oriented systems built around restricting/controlling/limiting/funnelling distributed naming, through a Proof of Work pipeline.




Most proponents of Proof of Work don't celebrate the use of energy. They celebrate the use cases that PoW enables. I don't know of any other way to trustlessly establish a consensus ordering, and I believe that being able to trustlessly establish a consensus ordering is very valuable, therefore I believe PoW is worth the expense.

These decisions get made on an open market, it's not like anybody is forced to put money into a wasteful system if they aren't getting value out of that system.


I personally don't feel like identity & naming want or need concensus ordering. I cited HIP & Wireguard, which don't require that total ordering.

The issues of control & power that come into play when we create distributed systems then make them all agree & work in the same way seems antithetical, to me, to the very idea of distributedness itself. I'm far more interested in systems that can range & explore possibility spaces untethered from any notion of root consensus. I would have us rely on informal weak trust models, friend of a friend, web of trust relations, rather than distributed-but-totalizing.

Thank your for your reply. I've talked some about my questions & what I want to explore, but I appreciate your comment a lot. Talking about what proof of work buys us, that it's there for total ordering, is I think a wonderful assessment, & even though it's not my priority or desire, I do think it buys a lot of interesting technical capabilities & assurances we may otherwise not be able to have.


Wireguard and HIP don't allow people to reserve human-chosen names in a single namespace and then resolve conflicts between users trying to reserve the same name. Solving that problem in a trustless decentralized way takes a blockchain-type system. (The problem is sometimes called "Zooko's Triangle". There weren't any known good solutions to it before Bitcoin happened.)


Right, for sure.

I'm not really in favor of single namespaces at this point. It's been wonderful having DNS as a reliable universal naming system, but if we're talking about trying to move to a more distributed model, I'm more interested in loosening the strings & exploring the space beyond a single namespace. Not just decentralized, as the article talks about, but also distributed. Embrace strong cryptographic identity, cryptographic names first, & leave human naming for additional layers, probably around web-of-trust models.

To me, trying to recreate DNS but with arbitrary distributed blockchain systems to allocate power instead of arbitrary icann registries doesn't seem that advantageous. I do like the idea of censorship resistance, of really owning a name, but the cost & viability of claiming & winning names via out Proof-of-Work'ing folks or buying currency seems daunting to me. The hurdles here scare me. And having one single namespace that can never expand & grow, where the good keeps getting carved out, by early players: that scarcity is unappealing. Single namespaces are consistent & that's powerful, but the digital is capable of so much less scarcity, albeit we humans only have limited experience trying to harness & direct such unlimited digital abandons.

There's plenty of good, interesting, valuable work on single namespaces that we can do, & this is probably a good, interesting, & valuable contribution (albeit one that leaves me with a lot of questions about participation & how the system functions). Restricting ourselves to the scarcity of a single namespace seems both sensible & practical, like a straightforward route towards adoptability, towards a more likely to be impactful protocol. But I do hope some of the wilder, less restrained options for how we do naming get a chance, are a part of our broader experiments in naming.


> Restricting ourselves to the scarcity of a single namespace seems both sensible & practical, like a straightforward route towards adoptability, towards a more likely to be impactful protocol. But I do hope some of the wilder, less restrained options for how we do naming get a chance, are a part of our broader experiments in naming.

Do you have some particular options in mind? Practically any example I am can come up with ends up devolving to a single namespace.

The least restrictive one I can recall is the Usenet namespace where the equivalent of a subdomain didn't — at least in principle — require permission from the domain "owner" to establish (eg. alt.comics.superman.dies.dies.dies). In practice this required coordination and consensus among various administrators for a new group to be established, and under various scenarios existing "owners" could object.


Overall I think letting people pick their own pet names & grow their personal popularity is the most important step we can take as an onlining civilization. It's also going to be total chaos, but I think we'd have to keep growing & iterating, via a lot of bad failing. Present the complexity, present the multitude of options, & surface historical data to show who favors & prefers what, over time.

A more practical take, I'd love a domain name system that only promised that, at some time, i was the owner of a certain DNS domain. If I stopped being the owner, I could still continue hosting my site, my identity would be cryptographically well known & certifiable, people would know who to come to (or maybe not, not a hard requirement for this scenario; like HIP they should know me when we talk & i do want to say it's me), even though the link, the "ownership" (renter-ship) of the domain i'd held had expired. This again is another huge host of interesting messy hard problems, and if someone does get my keys, i'm hosed, but i really like how it favors a very strong distributed identity over this pressing incessant badgering need DNS has to enforce consistency & singularness.


Hmm. So something like sending mail to a person at a street address? That is, even if you don't still live there (and even if there is no forwarding address), it's obvious who it is for (which isn't the current occupant).




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: