Nostr Web Client

You do hodl your own keys with Ark, and those keys can exit unilaterally after one block (comparable to onchain levels of sovereignty, pluse state-chain security before that moment, which is less than a DryjaPoon channel but more than you can say onchain). In very high fee scenarios, unilateral exit is economically feasible for most amounts in Ark, but only for very high amounts in DryjaPoon channels.

Running your node is a bit orthogonal. It's just that an Ark client node on top of a L1 node is a trivial thing, unlike a node to manage DryjaPoon channels.

Protecting your own privacy is something you can do well with SOME current DryjaPoon channel setups (with unannounced channels, blinded paths and Tor, for example), and that is still slightly worse for Arkade afaik right now. But it can get very good (see the original Ark paper with also tumbling and chaumian stuff included).

I agree with you that "90%, is very possible with LN". Today the LN is a universal, global, invoicing/hopping/swapping standard across many local security models with different security/usability tradeoffs: L1 (moon), ecash (minibits, fedi, mutiny, etc.), liquid (aqua, bull, breez, etc.), spark (WoS, Blitz, etc.), dryjapoon (phoenix, etc.). I think this complex variety will eventually collapse into a simpler number of primitives (namely: ecash for super small amounts, a mix of arks and spilman channels for average amounts, onchain for super big amounts).

Using your own brain is very likely orthogonal. 😉

Matt Corallo 1mo ago 💬 1

> You do hodl your own keys with Ark, and those keys can exit unilaterally after one block (comparable to onchain levels of sovereignty, pluse state-chain security before that moment, which is less than a DryjaPoon channel but more than you can say onchain).

Worth pointing out that Ark doesn’t materially “scale” without adding payment channels on top (ie Timeout Trees). Until that point, you’re really stuck with the statechain model, which has a highly-trusted third party, as you note.

> In very high fee scenarios, unilateral exit is economically feasible for most amounts in Ark, but only for very high amounts in DryjaPoon channels.

Huh? Ark is strictly more expensive to unilaterally exit than classic lightning.

> Running your node is a bit orthogonal. It's just that an Ark client node on top of a L1 node is a trivial thing, unlike a node to manage DryjaPoon channels.

Also huh? You’re comparing apples and oranges. Yes running a full routing node is complicated to manage, but having DryjaPoon channels in a wallet is not fundamentally complicated (see eg Phoenix). The second model (“client” to an LSP) is more comparable to Ark, and of course as mentioned above if you want scalable, trustless Ark you need payment channels anyway, so it’s strictly more complicated!

Reply to this note

Please Login to reply.

Discussion

Matt Corallo 1mo ago 💬 2

Oh of course also worth pointing out that for a mobile client to use an Ark it’s never going to be trustless without a soft fork as well. You can join a huge multisig if you’re a server and always-online, though (but you still need channels to make it scalable!)

Iain Cheyne 1mo ago 💬 1

Thanks. I didn't realise Ark needs CTV to be trustless on mobile.

Matt Corallo 1mo ago

It’s…. Complicated, but basically the way they’re working around needing CTV is to do a big multisig with everyone. That requires a decent amount of “being online at the right time”, which maybe you can get lucky with on mobile but probably not.

Giacomo Zucco 1mo ago 💬 1

Sorry for the delay. Nostr still sucks. :)

> Worth pointing out that Ark doesn’t materially “scale” without adding payment channels on top (ie Timeout Trees). Until that point, you’re really stuck with the statechain model, which has a highly-trusted third party, as you note.

I'd say you have the statechain model *until* you get the confirmation, then the model is closer to a DryjaPoon LSP than a statechain, until the expiration time. But yes, I agree the Ark model will be really interesting with channels on top, they are working on it since a while and it makes sense: https://blog.arklabs.xyz/bitcoin-virtual-channels/

> Huh? Ark is strictly more expensive to unilaterally exit than classic lightning.

I guess I'm playing with the word "unilateral" a bit here. I think that, realistically, the relevant thing is the exit against the coordinator, not against all the other users. The total cost of an unilateral exit in Ark is higher, but you can conceptually split it among more people in most failure modes. In a classic DryjaPoon, you share an UTxO among 2 users. That means that in best case scenarios (cooperative) the cost for opening/closing/resplicing would be X/2. In uncooperative scenarios, the cost for the exiting user is X. In an Ark with N users (ASP included), opening/closing/closing has a cost of about 10X, but that would be shared among N in cooperative cases, N-n in case n (most importantly including the ASP) are uncooperative.

> f course as mentioned above if you want scalable, trustless Ark you need payment channels anyway, so it’s strictly more complicated!

Good point.

Matt Corallo 1mo ago 💬 2

> I'd say you have the statechain model *until* you get the confirmation, then the model is closer to a DryjaPoon LSP than a statechain, until the expiration time. But yes, I agree the Ark model will be really interesting with channels on top, they are working on it since a while and it makes sense: https://blog.arklabs.xyz/bitcoin-virtual-channels/

Yea depends a lot on how refresh happens and how much you transact. If you transact daily and refreshes are only once a month you’re effective always in the statechain world (at least until payment channels on top of Ark!). This is kinda my base case. If you’re more of a receive-only/savings wallet then it’s less trusted. Good luck explaining that to a user 😅.

> I guess I'm playing with the word "unilateral" a bit here. I think that, realistically, the relevant thing is the exit against the coordinator, not against all the other users. The total cost of an unilateral exit in Ark is higher, but you can conceptually split it among more people in most failure modes. In a classic DryjaPoon, you share an UTxO among 2 users. That means that in best case scenarios (cooperative) the cost for opening/closing/resplicing would be X/2. In uncooperative scenarios, the cost for the exiting user is X. In an Ark with N users (ASP included), opening/closing/closing has a cost of about 10X, but that would be shared among N in cooperative cases, N-n in case n (most importantly including the ASP) are uncooperative.

No, LN is still always cheaper (modulo on-chain HTLC resolution). In Ark even if everyone is exiting and splitting the cost (which isn’t quite how it works but let’s assume), the cost to exit is one-two transactions per wallet (average of one tree transaction plus one transaction to resolve their vTXO. You could do the resolution later so maybe it’s one but depends a lot on the specific construction). In LN it’s always only one.

If you have any HTLCs pending you have to resolve them in LN, but I’m kinda assuming all the Ark things will mostly be used with HTLCs so you’d end up with the same in Ark.