Thread by @JoelKatz

@XrpMr and @MikeV_XRP : Just saw your question a few months ago about transaction fees on XRPL. To a first approximation, if there's X txns per ledger, you just have to pay more than X-1 txns. But here's the gory details:

Each XRPL server maintains an "open ledger" that it test applies txns to. This is how it determines if a txn seems likely to succeed. When the ledger closes, validators take the txns in this ledger as their initial proposal.

There are three ways to get into those open ledgers and thus in a validator's proposal. The first is if you are rejected in a previous consensus round. Those go in first.

This is important for sane system behavior and censorship resistance, but doesn't affect fees much. But the next step is the ledger is "filled". Txns received by the server are applied to the new open ledger in fee order right after each consensus round.

This is how most txns get in the ledger. The server applies txns, highest fee rate first. As it does so, a fee escalation algorithm demands higher and higher fees as the number of txns gets higher.

The "knee" in the curve is based on observed network behavior. So it won't demand excessive fees unless more txns are trying to get in than the network can handle.

Once no txn is willing to pay the escalated fee, filling stops. At this point, any txn received that is willing to pay the "open ledger fee" goes right in, otherwise it waits in the fee queue until the next ledger is "filled".

In all the time I've observed XRPL behavior, I've never seen a time when a 20 drop fee (double the base fee) wouldn't get you to the tip of the queue within 15 seconds. /thread