> because it breaks the assumption that coins don't expire

Technically, they can expire, if the client can see some chain reorganization. And it is something to think about: if the consensus will force miners to go back to some previous block height, and produce a stronger, alternative chain, then such testnet would automatically perform full chain reorganization, wired into its consensus rules. And then, coins would expire in a backward-compatible way, while also battle testing full chain reorganization (which may need testing anyway, for other reasons, like checkpoints).

> This will eventually overflow, but that seems fine for a testnet.

As far as I remember, there were some additional limits, introduced after Value Overflow Incident, for example that a single UTXO cannot hold more than 21 million coins: https://github.com/bitcoin/bitcoin/blob/master/src/consensus/tx_check.cpp#L29

```
// Check for negative or overflow output values (see CVE-2010-5139)
CAmount nValueOut = 0;
for (const auto& txout : tx.vout)
{
if (txout.nValue < 0)
return state.Invalid(TxValidationResult::TX_CONSENSUS, "bad-txns-vout-negative");
if (txout.nValue > MAX_MONEY)
return state.Invalid(TxValidationResult::TX_CONSENSUS, "bad-txns-vout-toolarge");
nValueOut += txout.nValue;
if (!MoneyRange(nValueOut))
return state.Invalid(TxValidationResult::TX_CONSENSUS, "bad-txns-txouttotal-toolarge");
}
```

Which means that in practice, instead of seeing huge or overflowed amounts in UTXOs, we will probably see a lot of repeated entries in the UTXO set, holding MAX_MONEY each.

pon., 28 kwi 2025 o 14:47 Sjors Provoost <sjors@sprovoost.nl> napisał(a):

Jameson Lopp wrote:

> Encoding an "end of life date" into testnets is actually an interesting idea worth discussing. As far as I'm aware it's never been done before on any network.

Keep in mind that testnet-specific code has to live right next to, even inside of, mainnet consensus code. We want the change to be as simple as possible, so as to not accidentally break mainnet.

Unless and until coin expiration is something we're seriously considering for mainnet, we'd rather not implement it for testnet.

This particular idea probably requires a lot of changes all over the place (consensus, mempool, wallet) because it breaks the assumption that coins don't expire.

Something I've proposed in person a few times, is to double the coins every halving. In terms of code, it boils down to changing GetBlockSubsidy:

CAmount nSubsidy = 50 * COIN;
// Subsidy is cut in half every 210,000 blocks which will occur approximately every 4 years.
If (consensusParams.inflation) {
// Except on testnet5
nSubsidy <<= halvings;
} else {
nSubsidy >>= halvings;
}

This will eventually overflow, but that seems fine for a testnet. Along with the timewarp fix, the network might even grind to a halt in 2106, long before we overflow 64 bit numbers.

Rust Bitcoin [0] currently refuses amounts above 21 million BTC, but they would have many years to fix that.

Strong inflation has been battle tested by governments around the world for millennia as a way to discourage saving.

- Sjors

[0] https://github.com/rust-bitcoin/rust-bitcoin/issues/4273