How would you write a ledger for micro payments without blockchain?

By using traditional finance combined with modern electronic payment processing (credit card) ledger models and not actually doing micropayments. I'd do micro accumulation and only ever pay out once a threshold is reached.

Many credit cards rewards programs do just this. You accumulate points when you spend, where a single point may be a fraction of a penny, and you can only claim the cash-back reward with a minimum of X-hundred or -thousand so points.

It's just too inefficient to actually pay out extremely small sums in a single transaction. Blockchain doesn't solve this - you'd still burn more fuel than it's worth.

Global state consistency doesn’t require “global consensus”.

In other words; agents paying each-other don’t need the expense of awaiting agreement or total-ordering of transactions, in order for the global ledger to maintain a zero balance.

This allows the per-transaction cost to approach zero: the cost of several back-and-forth trips between the nodes involved in the transaction, and the storage of a small number of commits out to a shared DHT (a few kilobytes of unique data stored across several hundred DHT nodes).

The actual cost of resources used is almost immeasurably small, so the “transaction cost” of the micropayments would probably be subsumed in the revenue model of the goods/services being paid for (ie. “too cheap to meter”).

Since no artificial global consensus or total ordering is required, there is no limit on transaction rate — it rises roughly linearly with node count.

The name of the underlying technology is Holochain. You can build and deploy on it now (it’s in beta).

It doesn’t have a “cryptocurrency” associated with it, but a project called Holo has one called HoloFuel (used to pay for hosting Holochain based “hApps”, for those who don’t wish to self host); it presently exists as a proxy — an ERC-20 token called HOT — until Holo goes live.

Many other projects are presently implementing many “cryptocurrencies” (mostly as “mutual credit” style asset-backed currencies), but some are implementing multi-asset ledgers and order-books to provide multi-currency and cross-chain DeFi, without the cost and complexity of using eg. the EVM to do so.

Take a look. It’s a breakthrough. May take a bit to wrap your head around.

I would take any database (eg Mongo, Maria, …) patch out the „delete“ queries, and run the DB in a confidential cloud like enclaive.cloud. Leveraging attestation and confidential compute, one can add a proof that the DB does not allow for erasure. Running the DB in a (kubernetes) cluster on multiple clouds would ensure the fault tolerance known from blockchain ledgers.

https://www.patreon.com/ is a solution. but I was wondering if there's a p2p way to deal with it.

Formance https://www.formance.com

You might want to look at 1990s micropayments work like Millicent and Peppercoin.

I suppose one must quibble with what one terms “blockchain.” The ledger part is fine, the distributed/decentralized part is inefficient.

Use hierarchical private keys (which can be delegated among sub-groups) and use a secure infrastructure of delegation (no strangers, only key holders.) with “trusted” delegation, delegate heads could still use redundancy for maintaining trust (triple redundant recurrent results.)

Ledger of a block chain, without the waste or 51%+ attack vulnerability.

A private blockchain.