Manual transaction verification

[matiwinnetou]

Why

On Ethereum smart contracts offer ability to store global state. In Cardano due to favouring local state and offchain transaction evaluation developers do not have access to global state in the contract.

In some cases, however, access to global state is badly needed to avoid workarounds.

What

We would like to get a feature where each transaction can be manually verified by each head participant. Today verification / validation of transactions is automatic and there is no way to intercept this process and disagree with validating a certain transaction.

How

Hydra applications running in off-chain infra, could have ability to register via e.g. websockets as a manual hook to hydra-node. Each time hydra is about to validate transaction, transaction will be serialisation, hook will be triggered and off-chain app will have ability to act on this transaction.

Such a plugin and hook typically would be disabled and would need to extra enabled in configuration as it applies only to some class of applications.

Hook would act as an "off chain validator", receiving transaction, redeemer and script context all serialised.

[ltouro]

This seems more like an application level concern than a protocol level one. Would not be better to make 'manual confirmation' a transaction itself? There is all sort of contract patterns that would allow you to do that, like on-chain multi-sig with support to partially signed transactions.

[ch1bo]

I interpret this idea as essentially asking for a different ledger. One that runs external to the hydra-node and is given the same information as the Cardano ledger used on the L2 right now. At the very least it would mean to extend the ledger rules using this custom hook code. One question here: further constraining otherwise valid transactions is what we need, right?

Hook would act as an "off chain validator", receiving transaction, redeemer and script context all serialised.

This is indicating that your idea would still be EUTxO based and have a similar preparation / context be done like it is right now given to scripts. Why can the computation in question (an example would be good) not be done "on-chain", on the L2 already?