> > Hard fork in an ultramassive premine, as large as possible but what stays
> > with existing value overflow logic. (so maybe an additional 21 million
> > testnet btc?).

The existing logic gives errors if:

  * a single input of a tx (ie a coin in the utxo set), or the sum of inputs to
    a txn, is outside the range 0-21M (bad-txns-inputvalues-outofrange)

  * a single output of a tx is outside the range 0-21M
    (bad-txns-vout-negative or bad-txns-vout-toolarge)

  * the sum of the outputs of a single tx is outside the range 0-21M
    (bad-txns-txouttotal-toolarge)

  * the fee paid by a single tx is outside the range 0-21M
    (bad-txns-fee-outofrange)

  * a block's fees go outside the range 0-21M
    (bad-txns-accumulated-fee-outofrange)

Keeping the total supply under 21M seems nicer than having txs that
spend real utxos be able to hit these errors (eg, by combining
the premine utxo at 21M with a coinbase reward of 50 and hitting
bad-txns-inputvalues-outofrange).

That's pretty easy to achieve: just have the initial premine be half the
supply (eg), and also cut the halvening time in half (so 10.5M premine,
105,000 blocks in a halving). Or you could have halvenings every 6 months
(26250 blocks), and have an 18.375M premine, or whatever.

You could also consider premining (almost) the entire supply, and have
the block reward be entirely fees (almost) immediately after that, but I
think there's value in making it possible to obtain coins for testing in
a permissionless, anonymous and relatively low-latency manner, for which
PoW is great. Might also be annoying for empty blocks to pay a reward of
exactly 0, so if miners included their address in the coinbase tx like
normal, they'd be creating a 0 valued utxo, and probably never spend it.

I had a quick poke at what code to allow for chains with premines might
look like here:

  https://github.com/ajtowns/bitcoin/commits/202505-premine/

About 11 lines of code to implement the logic.

If this approach made the testnet difficulty reset logic obsolete
(ie, a testnet with just PoW and a premine turns out to work fine),
that would drop 14 lines of code for the fPowAllowMinDifficultyBlocks
and enforce_BIP94 logic. Presumably a PoW-only testnet could also have
its min-difficulty bumped from 1 to 65536 or more, since it seems like
a single Bitaxe can still maintain the chain at that difficulty.

The idea of this approach is that when establishing a premined testnet,
you would:

 a) first define the chain, with a new genesis, etc; then set
    nSubsidyHalvingInterval=105000 and premine=10'500'000*COIN or similar,
    but leave premine_block_hash=0

 b) build the node software, and mine block 1 to the premine address.

 c) set premine_block_hash to block 1's hash. publish the code with the
    genesis block and block 1 hash, so that the public can mine as of
    block 2.

 d) once 100 blocks have been mined, split the premine up amongst
    devs, faucets, wallet maintainers, user groups, a managed endowment
    for future testers, whatever.

On Fri, May 09, 2025 at 03:07:48PM +0200, Garlo Nicon wrote:
> Why hard-forking anything? The starting difficulty is set to 1, and it
> raises to 4 almost instantly, when testnet creators are mining the first
> coins. Which means, that difficulty 1 is ridiculously easy to work with,
> when you have any ASICs. If you combine it with the idea of fake
> timestamps,

It's not the number of blocks, but the cumulative work that matters, so to
have a soft reset of testnet3 or testnet4 you'd need to apply more hashing
for the new chain than the existing chains have already received. That's
a fair amount of "wasted" hash: I think mining a more-work chain than
testnet4 would require about the same amount of hash that it would take
to mine ~13 mainnet blocks at the current difficulty, so you'd be giving
up about $4M USD in mainnet block rewards to do it.

In any event, a hard fork is "necessary", as otherwise whenever it
takes 20 minutes or more to find a block, old clients will expect a
lower difficulty than new clients do, so the two wouldn't be compatible
with each other. You could do various things to work around that, but
that's a lot of coding time that could be better spent on improving
things relevant to mainnet, and if you're resetting the chain anyway,
there's not much advantage to it.

> then you can produce a really long initial chain, which will
> start in 2009, and up to 2025, it will produce almost the same amount of
> blocks as mainnet.

A soft fork of testnet3 would start 3rd Feb 2011 (leading to about 750k
blocks vs mainnet's ~900k), and a soft fork of testnet4 would start at
4th May 2024 (leading to about 54k blocks). (These are the timestamps
of the respective genesis blocks)

A disadvantage of doing a premine that way is that users of the chain
need to download and validate thousands of blocks and deal with an equal
number of utxos just to establish the premine; doing that in a single
block with a single utxo (or one utxo for each recipient of the premine)
is quite a bit more efficient.

> Which means, that instead of "premine", you can use "ninja-mine", and
> achieve pretty much the same end result.

I think in general usage "premine" covers both those approaches -- any
time the creator(s) of a chain gets the opportunity to claim/distribute
coins prior to the general public being able to mint new coins by mining
blocks, that's a premine.

Cheers,
aj