Yeah my comment was all over the place, but I hope not too much to be totally worthless.
PoA doesn’t mean the validators can change history, at least unless the network is specifically designed for that – which most aren’t, although I’d argue there’s potential use cases for allowing to eg. “undo” transactions, like what your bank does if your credit card number gets stolen.
Re. partitionable chains, it’d be fun to think about how to manage transactions when some nodes are potentially light years away. We already have the “interplanetary file system” after all 😄
jet@hackertalks.com 1 year ago
Well we kind of already have historical examples of partitionable ledgers. It’s all about the merge. So historical documents written in far off offices merge or central offices. And they just kind of ignored conflict effects.
So if you have a very partitionable environment and we are using a distributed blockchain we might have to do something like record authority moving between partitions. That could be really interesting
interolivary@beehaw.org 1 year ago
Eventual consistency would be really eventual, heh.
And if the network is generally partitioned (or DAG-like I guess?), how do you handle eg. someone hopping on a (slower-than-light! I don’t believe in that FTL nonsense) ship and going from eg. Earth to live on Alpha Centauri? Do they have to bring a part of the DAG with them (or some sort of zk proof of it anyhow) in some form, so that it can be “transplanted” into the consensus on the other end when they arrive?
jet@hackertalks.com 1 year ago
I can only imagine in such an environment you would have packetized network updates. You wouldn’t try to run a globally consistent ledger. You would bake in the partition network. And if you know somebody’s going to transit from network aid to network b you might sign something you might do a key you might give some sort of authority for this record to now get updated in a different partition. And if that traveled with a human all the better.
I imagine the partitions would also keep best effort consistent copies of other partitions, but they wouldn’t consider them you know up to date for any logistical purpose. So if you had a record that dealt with partition a but you were in partition b you would leverage as much data in partition be as possible and send an update record to a who would be the authority to do the thing and then send you the result.
But that’s a very very very naive approach. I’m sure we could come up with something more interesting. A distributed eventually consistent ledger with very sparse updates. Could get interesting
interolivary@beehaw.org 1 year ago
And this is exactly what I meant when I said that this sort of interstellar networking stuff is super fun to think about 😄