The Scalability Question
DAGs don't scale well. Discreet has a solution.
Directed Acyclic Graphs have a well-known scaling issue. Discreet believes that DAGs have distinct advantages over regular blockchains that make them preferable for the kind of network we’re building, but they do begin to exhibit issues under heavy load. That’s not wrong.
Discreet was of course aware of this limitation and was designed to address it in part through our extension of Aleph technology. This is why the fair consensus model we’re proposing would only ever include 100 to 150 nodes at any given time, pulled randomly from a larger pool of approved candidate nodes in a rotating committee. The nodes would be elected into a Discreet Consensus Committee for a set number of rounds (called an “epoch”).
After an arbitrary number of rounds, the randomness beacons coalesce around one node to switch out and replace with another from the pool of potential consensus nodes (again, selected via the randomness beacon). Nodes that want to be considered for that pool would perform a small proof-of-work to ensure Sybil-resistance (so each node would work on its own PoW validated by the other nodes before being accepted into the pool of candidates). Note: This PoW would have a difficulty low enough that the computing problems would not become too resource-intensive, allowing regular people on decent modern personal computers to participate, using a method like Monero’s RandomX, which is ASIC-resistant.
In this manner, the consensus process remains fair and just as decentralized as any typical proof-of-work mining, but with the added speed advantage of proof-of-stake on a DAG.
Being chosen as part of the Discreet Consensus Committee is essentially the same as finding the hash with however many leading 0's on the Bitcoin network, except in our case rather than that person deciding how the block is constructed, up to 150 nodes would have to agree on how the block will be minted. (This is so that if there is someone trying to mint fraudulent blocks, we could assume more honest nodes than not, and follow majority consensus.)
There will be a push/pull balance between centralization and decentralization, at least at first, as Discreet builds a network capable of the power and speeds we anticipate being possible. At first, because of the limitations inherent in DAGs, Discreet may be limited to around 2000 transactions per second. This will be sufficient until the project can grow the Discreet Compute Network of Trustless Autonomous Compute Nodes (TACNs).
It is important to understand that dApps on Discreet’s network will function using Discreet’s paranets, like the dApps on the Polkadot network’s parachains, or Solana’s similar solution. In this way, some dApps may run on smaller side chains, and others will run on the main Discreet network. These multiple flavours of on-chain and off-chain paranets will self-select based on whether individual computations are needed or if an entire network of services is required.
Discreet will support a modular mechanism similar to Ethereum’s EVM, where select TACNs will be able to extend functionality by running opcodes from other developers. (Discreet paranets will include full support for zk-SNARKS, for example.)
If anyone has any questions about any of this, you are welcome to ask our developers in Telegram or through the form at the bottom of our FAQ page. More details on Discreet’s network and consensus methods will be made available in the coming months in a yellow paper which will explain these solutions in greater mathematical detail.