Hmm, depends if this growth is ephemeral or not. If it's cleared away somehow, PoW style, it could be tenable,
The trouble is, the stake signing needs to be verifiable, otherwise you'd need to have a consensus for forming a consensus!
Topic: Instant transactions with a stream of work -- Quanta (Read 663 times)
The trouble is, the stake signing needs to be verifiable, otherwise you'd need to have a consensus for forming a consensus!
Interesting - I've been thinking along similar lines recently, especially the stream of transactions. I was leaning more towards having each node 'sign' a transaction with it's stake as they get relayed between nodes, such that you get a fast forming, POS style consensus on each transaction as it passes between nodes. The only (and significant) problem with this is that it causes transactions to grow as they pass between nodes.
Hmm, depends if this growth is ephemeral or not. If it's cleared away somehow, PoW style, it could be tenable, though there might be trouble making the PoS mean anything to the miner; it would may necessitate everyone synchronising their mempools, to some degree at least, or including the PoS stuff in the block, though that would defeat the purpose. Hmm, indeed.
Last year I was working on a new sort of blockchain that would enable instant transactions. I'm dumping some info here so it's searchable and maybe inspires someone or w/e.
The basic idea is that each TX has a small PoW attached and is 'mined' instantly. Larger miners (who are in it for the money) mine TXs with *large* PoWs and point back to other 'work heavy' blocks but also link back to these 'lighter' txs from regular joes.
The basic idea is that each TX has a small PoW attached and is 'mined' instantly. Larger miners (who are in it for the money) mine TXs with *large* PoWs and point back to other 'work heavy' blocks but also link back to these 'lighter' txs from regular joes.
Interesting - I've been thinking along similar lines recently, especially the stream of transactions. I was leaning more towards having each node 'sign' a transaction with it's stake as they get relayed between nodes, such that you get a fast forming, POS style consensus on each transaction as it passes between nodes. The only (and significant) problem with this is that it causes transactions to grow as they pass between nodes.
Last year I was working on a new sort of blockchain that would enable instant transactions. I'm dumping some info here so it's searchable and maybe inspires someone or w/e.
The basic idea is that each TX has a small PoW attached and is 'mined' instantly. Larger miners (who are in it for the money) mine TXs with *large* PoWs and point back to other 'work heavy' blocks but also link back to these 'lighter' txs from regular joes.
The blockchain is actually a DAG instead of a linked list.
One challenge was ordering txs, I think I've come up with a pretty good solution that's included in the source here: https://github.com/XertroV/quanta-test/blob/master/quanta.py#L204. Basically it recurses down the heaviest path until it reaches a common point between all paths, then it recurses down the second heaviest path till it hits a block in the first path, and so on down all paths. In this way it can operate as a DAG, has a deterministic ordering, and not worry about people inserting blocks into history or anything like that (though there might be a DoS angle here on computationally heavy re-orgs).
It relies on treating the pool of workers available as producing a *stream* of work, rather than the discrete blocks that we're used to. By treating it as a stream you get (nearly) infinite granularity, enabling near instant TXs. You still have to wait an hour for a good confirmation, though (like all networks).
One downside is all TXs have lots of metadata about links so they can be 500 bytes for a simple TX.
Source code is here: https://github.com/XertroV/quanta-test
I can't remember if it works or not when you run it.
The basic idea is that each TX has a small PoW attached and is 'mined' instantly. Larger miners (who are in it for the money) mine TXs with *large* PoWs and point back to other 'work heavy' blocks but also link back to these 'lighter' txs from regular joes.
The blockchain is actually a DAG instead of a linked list.
One challenge was ordering txs, I think I've come up with a pretty good solution that's included in the source here: https://github.com/XertroV/quanta-test/blob/master/quanta.py#L204. Basically it recurses down the heaviest path until it reaches a common point between all paths, then it recurses down the second heaviest path till it hits a block in the first path, and so on down all paths. In this way it can operate as a DAG, has a deterministic ordering, and not worry about people inserting blocks into history or anything like that (though there might be a DoS angle here on computationally heavy re-orgs).
It relies on treating the pool of workers available as producing a *stream* of work, rather than the discrete blocks that we're used to. By treating it as a stream you get (nearly) infinite granularity, enabling near instant TXs. You still have to wait an hour for a good confirmation, though (like all networks).
One downside is all TXs have lots of metadata about links so they can be 500 bytes for a simple TX.
Source code is here: https://github.com/XertroV/quanta-test
I can't remember if it works or not when you run it.
Jump to: