Mempool Disccusion

[danielSanchezQ]

I'm creating this to discuss about the current mempool implementation.

We know that at some point we would want to keep the transactions stored in some kind of decentralized storage (project CODEX). For not we are making the mempool hold the pending txs which makes sense, but for the committed ones we could probably move them to some storage that can be later replaced by the original idea.

[zeegomo]

If we decide that blocks only contain transactions hashes, then we need a way to retrieve transactions. It's not too different from how a node that joins the network later has to download blocks actually.

I think it could be valuable to have committed transactions in a coldish storage and conflate storage of pending transactions and transactions in blocks not yet committed, as we might have to access both

[danielSanchezQ]

If we decide that blocks only contain transactions hashes, then we need a way to retrieve transactions. It's not too different from how a node that joins the network later has to download blocks actually.

I think it could be valuable to have committed transactions in a coldish storage and conflate storage of pending transactions and transactions in blocks not yet committed, as we might have to access both

That way we make the mempool work only over the transactions ids right? So if retrieving a transaction it should be from this storage?

[alvatar]

The mempool is subject to further research at this point. We cannot make an informed decision yet. For now, the simplest mempool will do, and then we will work on this from a research perspective. To clarify a few points:

• Codex cannot be used for the mempool. It can be used for archival nodes, but that's a different use case.
• The mempool will have a focus on privacy, likely being threshold-encrypted.
• Ideally, the mempool guarantees data availability (ie like Narwhal), and the consensus will operate on the hashes of the mempool.