[mattdesl]

As usual, “just trust Amazon or Google” which is centrally owned, and not peer-to-peer (distributed/decentralized), and not an answer to my question.

> You're mentioning a .eth domain name being bought with cryptocurrency as an example, which is entirely circular. "Hurr durr, betcha can't swap one blockchain thing (.eth domain) for another blockchain thing (cryptocurrency tokens) without using a blockchain" isn't as strong an argument as you think it is. If we were talking about a .com domain name, no blockchain in the world will help you with that transaction.

Your argument feels in bad faith, but I’ll bite: “.eth” and ENS is a valuable construct for those transacting in the network. These assets do have clear market value, even if you personally feel they shouldn’t.

[RandomLensman]

Nothing is stopping a peer-to-peer network from having a central database administered by the a randomly selected group of peers, for example.

The bigger issues in pure decentralized and distributed networks in finance are around KYC/CFT etc. Who ensures compliance if there is no control about flow of funds, for example.

[mattdesl]

What you are suggesting sounds like a consensus mechanism. How do you safely choose which peers to assign this responsibility? How do you ensure it is resistant to a Sybil attack?

Suddenly the answers begin to look a lot like Proof of Work or Proof of Stake.

[RandomLensman]

The answer could or could not look like that. In practice people have done this many times throughout history totally without blockchains (or computers) for that matter (e.g., money transfer systems, exchanges, etc.).

[mattdesl]

If this has been done many times in a way that solves my original question (decentralized escrow), there should be some concrete examples you can point to besides blockchain/crypto systems.

[RandomLensman]

Obviously, non-digital assets, but does the Hawala system fit your analog version for decentralized escrow? Early stock exchanges were sometimes created to get around existing third parties ("auctioneers") and allow the brokers to directly transact p2p. What about a credit coop (it is central, but it is also owned by all the users)?

Edit: at it's most basic, reliable coinage was kind of way to create reliable p2p abilities without risk of "double spend". Once the coins were out there, central authority didn't matter so much, i.e., "good" coins were used fair and wide beyond the coining state (e.g., Athenian tetradrachma). Funny add. in some areas people actually allow temporary double spend (so that can be another solution)...

[mattdesl]

Perhaps, but my original question was around digital assets across the internet (and ideally on a global scale). If you are trading a plush toy for a coconut the "atomic swap" could be done with each party's two hands.

[trinovantes]

So how do you coordinate this so called decentralized transaction? Over the internet through SSL certs that are centrally signed? You're still putting blind trust in something.

[mattdesl]

Transactions are broadcast with RPC. Once accepted and written to the ledger (ie: after a number of confirmations), you can verify the state of the transaction via your own local node.

It is impossible to completely remove the need for trust. We trust that our computers work as expected, that our modems and routers are not compromised, that RPC endpoints and software is running as expected, that the internet infrastructure in our country is sending messages correctly.

The blockchain isn’t a catch-all solution to our need to trust things in life. But it does allow us to, say, record and alter global state without placing it in the control of a single centralized intermediary.

[trinovantes]

You trust the private companies/individuals making your hardware and protecting your communications but suddenly trusting a private escrow is heresy? Seems like a weird double standard.

> record and alter global state without placing it in the control of a single centralized intermediary

Paxos solved this in the 90s

[mattdesl]

A centralized escrow is not heresy; it involves a different set of trade-offs. In some cases the decentralized escrow might be more appealing. To go back to my original example of a domain transfer, the exchange can occur in a matter of seconds or minutes within a blockchain, rather than 1 to 20 business days with escrow.com.

Never heard of Paxos, if it could achieve the same problems I've outlined earlier, I'd be curious to see it implemented.

[trinovantes]

The 20 days delay is a problem with escrow.com specifically. Domain sales are a trivially automated process that should be instant e.g. literally any domain registrar.

Paxos family of algorithms solves distributed state replication. It is the backbone of the database engines that already power most of the internet.

https://en.wikipedia.org/wiki/Paxos_(computer_science)

[mattdesl]

Namecheap: 10% commission, only works with Namecheap-registered domains, the exchange may take up to 96 hours, and then 5 days later you can withdraw these funds to PayPal (which will incur additional fees).

Compare this to, say, Tezos domains: exchange and transfer of funds settled in ~30 seconds, without any need for currency conversion, across any ".tez" domain in the network, 2.5% commission (or 0% via custom contract), no private data shared with registrar, and very low transaction fees.

Looking at Paxos: it is permissioned, lacks Sybil protection, uses leader-based rather than peer-to-peer data replication, and seems limited in how many nodes it can support. This isn't to say it's useless, but it clearly aims to solve a different set of problems than Nakamoto's consensus mechanism (and, more generally, cryptocurrency networks).