Lightning network hub centralization

Hey guys - First of all, very impressed by the work that you are doing. Interesting topic!

I had a query. Have you read about the centralization concerns in Lightning Network? [1] Is this something relevant to Marlin? Could the emergence of hubs affect the security of Marlin?

[1] https://www.trustnodes.com/2020/02/14/lightning-network-increasingly-centralized-removing-hubs-leads-to-collapse-says-blockstream-study

TL;DR: There’s little to no centralization pressure on the network layer especially compared to LN.

Glossing over a lot of nuance, most centralization pressure in LN comes from the fact that LN channels are expensive. Expensive to open and expensive to maintain.

  • You need to make on-chain transactions to open channels which are expensive. Not to mention what such transactions need to cost in the future if transaction fees are to replace mining rewards (estimates are in the thousands of dollars in the same article). In contrast, you simply need to allocate file descriptors to open network connections.

  • LN channels decay over time. Most commerce is asymmetric. I only receive money from my employer, almost never pay them any. I only pay money to my grocery store, almost never receive any. This means that channels are inevitably going to run out on one side and need to be replenished. On-chain transactions again! Network connections on the other hand don’t really “decay”, their capacity is available 24/7 without fail (modulo oversubscription).

While LN might be much cheaper than on-chain transactions, it is still much more expensive than simple packet networks. And that’s not a knock on LN, it does a whole lot more than just moving packets here and there which drives up the cost. These are just reasons as to why they aren’t really comparable.

Another relevant thread: https://twitter.com/Snyke/status/1228733328179769344

Some of the arguments apply as-is here as well.