RE: Simple Side-chaining for Graphene-based Blockchains (BitShares/PeerPlays/Steem)
That's actually a brilliant idea. But requires the witnesses of the BITSHARES blockchain to know about the PEERPLAYS blockchain because otherwise they couldn't be sure that the blockheader and the signer is indeed a legit block producer on the other chain.
My goal was to have a means of bringing arbitrary many "side-chain"-groups to a blockchain without involving the actual consensus mechanism for the very simple reasons of maintaining scalability.
Let's assume you had a "master-blockchain" and wanted to add other blockchains as, erm, .. special-feature plugins blockchains, then you would want to be able to handle arbitrary many. For that reasons I think you should keep the consensus of each chain separately .. That's also good for robustness of the "master-blockchain" and the plugin-chains because they can all act independently ..
If we could come up with a proof that that can be verified without syncing the other blockchain, then your strategy is a winner!
We could use the committee for bootstrapping sidechains. A sidechain consists of at least (genesis hash, initial block signer's pubkeys, top-level asset name). Based on the genesis hash and the initial block signers, a chain of transferrable proofs can be generated. The transferrable proofs would have to continuously update the active block signing keys, and it'd have to produce a continuous chain of block header signatures. A transferrable proof is only accepted by the target chain after a sufficient number of successive blocks signed by 2/3 of the active witnesses has been received.
One tricky part is that Graphene currently allows to swap a large number (all?) of active witnesses (or their block signing keys) at once.
Storing block headers of several chains would create some bloat, but I think that'd be worth it.
Good to see that you came upt with a similar idea of what I have in mind as well .. at least long term. This article, however, focuses on a simple solution at the cost of trustlessnesa, of course.
I would love to see something like what you just explained potentially even with improvements to graphene, but in the meantime (lets assume we are going to need this soon), i think the approach above works sufficiently well.
Interestingly, though, a POW scheme could do better here because the difficulty is easier to track and stays constant for a certain period. That makes proofing a statemwnt in another chain easier (assuming sufficiently high difficulty)
: )