We should cache RA collateral (PCKs, TCB infos, ...) in an on-chain registry in the consensus layer that can be looked up in case Intel services are not available. Most likely this would be a separate consensus service (ra), built in a modular way where Intel SGX is just one of the supported technologies.
Then a PCS client could be built that uses this service as a fallback in case the node's usual PCS endpoint is not available.
Also optimize size of node registration transactions by allowing references to on-chain collateral.
We should cache RA collateral (PCKs, TCB infos, ...) in an on-chain registry in the consensus layer that can be looked up in case Intel services are not available. Most likely this would be a separate consensus service (
ra
), built in a modular way where Intel SGX is just one of the supported technologies.Then a PCS client could be built that uses this service as a fallback in case the node's usual PCS endpoint is not available.
Also optimize size of node registration transactions by allowing references to on-chain collateral.