As of today, 25 October, the Plenipo repository is public. Anyone can read the library, the specification, the test vectors and the sandbox, copy them and criticise them, and the Apache 2.0 licence lets them do whatever they like with it, including a product that competes with mine.
I know that to many people this will look like a gift made too early, by a project that is still young and has no customers. I believe there was no other way to start something like this, and I will try to explain why.
Whoever verifies must be able to read
Plenipo sits at a delicate point, the one where a bank decides whether to let a piece of software move money on behalf of a person. The person approves a mandate from their own EUDI Wallet, stating what the agent may do, towards whom, up to which amount and until when, and the bank checks that mandate before executing anything.
Whoever runs that check takes on a responsibility, and nobody should take it on the basis of code they cannot open. That holds for the institution, for the regulator who will one day ask it to account for a decision, and for the citizen, who has the right to know how what they signed is being read.
A public standard is not completed by a private piece
The EUDI Wallet comes from a European regulation and from specifications anyone can download, and the Commission's reference wallet is itself free software. Everything underneath Plenipo is open by political choice even before technical choice, because the digital identity of four hundred and fifty million people could not depend on one supplier's catalogue.
Placing a closed component exactly where that identity is used to delegate would have seemed a contradiction to me. Delegation is the new part, the one with the fewest certainties, and that is where the most eyes are needed.
Delegation without an owner
We are moving towards a world where agents book, buy and pay in our place, and today the only way to let them is to hand over passwords, keys or card details, which means full powers with no limits and no expiry. I strongly believe in the need to replace those full powers with a mandate, and I believe a mandate only makes sense if the same document is valid at any bank and with any agent.
If it worked only inside the product of whoever invented it, we would have rebuilt the dependency we wanted to remove, under a more elegant name.
The limits of an alpha
Working on LocalAI taught me that open code ages better, because the mistakes are found by the people who use it before the people who wrote it. It works the other way round too: opening up means showing your limits, and Plenipo has plenty.
It is an alpha. The cryptography is real, but no external firm has verified it yet, today's test passed with a single wallet, the one from walt.id, and nobody should use it with real money or real data. How a project like this sustains itself over time is a question I do not have a complete answer to yet, and I preferred to publish the code before having one rather than the other way round.
In the coming months what will count is who decides to read it. A bank trying to break it, a developer opening an issue, someone using it for a case I had not thought of: that is where I will learn whether this is the right road, far more than from any presentation.