It's been reported that the double negative in rule 2 is confusing people (including myself, on re-read). This PR revises out both negatives and, I believe, significantly improves clarity.
For reference, this document was based on the implementation in Bitcoin Core 0.12.0, which includes these lines:
for (unsigned int j = 0; j < tx.vin.size(); j++)
{
// We don't want to accept replacements that require low
// feerate junk to be mined first. Ideally we'd keep track of
// the ancestor feerates and make the decision based on that,
// but for now requiring all new inputs to be confirmed works.
if (!setConflictsParents.count(tx.vin[j].prevout.hash))
{
// Rather than check the UTXO set - potentially expensive -
// it's cheaper to just check if the new input refers to a
// tx that's in the mempool.
if (pool.mapTx.find(tx.vin[j].prevout.hash) != pool.mapTx.end())
return state.DoS(0, error("AcceptToMemoryPool: replacement %s adds unconfirmed input, idx %d",
hash.ToString(), j),
REJECT_NONSTANDARD, "replacement-adds-unconfirmed");
}
Changing this was suggested by @jnewbery, who also reviewed the code in master with me (which is the same but is moved to validation.cpp).
CC: @petertodd
Request to BIP editor @luke-jr: I think your usually policy is to merge policy-acceptable changes to BIPs solely based on the requests of their authors. As a co-author, I request this be left open for at least one week (December 13th) to give @petertodd a chance to comment if he wants.
