4 ms·

There is a big difference between your money sitting in a single file under the control of a single party and your money sitting in three files under the contro

There is a big difference between your money sitting in a single file under the control of a single party and your money sitting in three files under the control of three parties.

tomc1985 7y ago
I'm not so sure. Sure you can prevent one party changing the data in the DB but the decentralized model seems to carry a lot more risk. I can't wait for the day when todays encryption can be cracked by tomorrow's commodity hardware...
- EthanHeilman 7y ago
  >Sure you can prevent one party changing the data in the DB but the decentralized model seems to carry a lot more risk. Lets say you have 7 permissioned nodes that are running a BFT protocol to establish a shared DB. I'd be interested to hear why?
  - tomc1985 7y ago
    While I was originally thinking about the finance space and not the technicals of shared DBs themselves, I feel like N nodes introduces a weakest-link problem where one trusted actor subverts the rest. A hypothetically perfect BFT protocol would prevent this, I suppose, but I know nothing about them or their current state today. Are the resources guarding those seven nodes equally proportioned? How hard is it for a bad actor to potentially subvert one and cause it to pass bad-but-trusted data? Can these nodes be switched with other, potentially less-secure or hostile nodes? How are additional nodes added to the network; is there a credentialing process that could potentially be attacked? Who maintains each node, what is their level of sophistication, resourcing, and control?
    - EthanHeilman 7y ago
      BFT protocols are typically secure if 2/3 of the nodes are honest. >Are the resources guarding those seven nodes equally proportioned? How hard is it for a bad actor to potentially subvert one and cause it to pass bad-but-trusted data? Can these nodes be switched with other, potentially less-secure or hostile nodes? How are additional nodes added to the network; is there a credentialing process that could potentially be attacked? These are good questions. I believe in Libra human beings decide which organizations are allowed to run a node and they build a fixed list of say 100 or so nodes one from each organization. >Who maintains each node, what is their level of sophistication, resourcing, and control? You don't want amateurs running nodes. This is normally where such proposals fail because if you don't have enough clout early on you won't get good node operators and this poisons the security argument. From my perspective Libra is the first team which has overcome this hurdle. Then again hopefully everyone doesn't just run these nodes on AWS.

There is a big difference between your money sitting in a single file under the control of a single party and your money sitting in three files under the control of three parties.

tomc1985 7y ago

I'm not so sure. Sure you can prevent one party changing the data in the DB but the decentralized model seems to carry a lot more risk. I can't wait for the day when todays encryption can be cracked by tomorrow's commodity hardware...

EthanHeilman 7y ago
>Sure you can prevent one party changing the data in the DB but the decentralized model seems to carry a lot more risk. Lets say you have 7 permissioned nodes that are running a BFT protocol to establish a shared DB. I'd be interested to hear why?
- tomc1985 7y ago
  While I was originally thinking about the finance space and not the technicals of shared DBs themselves, I feel like N nodes introduces a weakest-link problem where one trusted actor subverts the rest. A hypothetically perfect BFT protocol would prevent this, I suppose, but I know nothing about them or their current state today. Are the resources guarding those seven nodes equally proportioned? How hard is it for a bad actor to potentially subvert one and cause it to pass bad-but-trusted data? Can these nodes be switched with other, potentially less-secure or hostile nodes? How are additional nodes added to the network; is there a credentialing process that could potentially be attacked? Who maintains each node, what is their level of sophistication, resourcing, and control?
  - EthanHeilman 7y ago
    BFT protocols are typically secure if 2/3 of the nodes are honest. >Are the resources guarding those seven nodes equally proportioned? How hard is it for a bad actor to potentially subvert one and cause it to pass bad-but-trusted data? Can these nodes be switched with other, potentially less-secure or hostile nodes? How are additional nodes added to the network; is there a credentialing process that could potentially be attacked? These are good questions. I believe in Libra human beings decide which organizations are allowed to run a node and they build a fixed list of say 100 or so nodes one from each organization. >Who maintains each node, what is their level of sophistication, resourcing, and control? You don't want amateurs running nodes. This is normally where such proposals fail because if you don't have enough clout early on you won't get good node operators and this poisons the security argument. From my perspective Libra is the first team which has overcome this hurdle. Then again hopefully everyone doesn't just run these nodes on AWS.