End-to-end demo: peering relay + policy-vs-consensus mining

[Mehdalon]

End-to-end demo: peering relay + policy-vs-consensus mining

Two local regtest demos of PR #3 actually working end-to-end, beyond the unit/functional/sanitizer/fuzz testing already covered in the PR thread. Both are deterministic, zero real money, no public network — proving the mechanics before anyone touches mainnet.

1. A large, dusty transaction relays over a real dog connection

Full writeup with the actual evidence and the recovered image: Mehdalon/dogmode-relay-demo

Short version: forced an actual dog-type connection between two nodes (same RPC the PR's own p2p_dogmode_peering.py uses), broadcast a 3,877,833 WU transaction — 99.4% of the actual 3,900,000 WU ceiling, 9.7x the old 400,000 WU standardness cap — carrying a 1-satoshi dust output and an image embedded in OP_RETURN, and confirmed it relayed to the second node solely over that connection — then extracted the image back out of the second node's own mempool copy and verified it byte-for-byte (SHA256 match, still opens as a valid photo).

2. Standardness is a relay courtesy, not a consensus rule

Second demo, aimed at a different question: if a transaction can't relay under default policy, can a cooperating miner get it confirmed anyway?

Two nodes: one running PR #3, one built from genuinely unmodified stock v31.1 (verified zero diff on the relevant files). Connected normally (no dog-peering involved this time).

• testmempoolaccept on the PR#3 node: allowed: true. Same call, same transaction, on the stock node: allowed: false, reject-reason: "tx-size". After broadcast, the stock node's mempool stayed empty — confirmed the relay wall directly rather than assuming it.
• Mined the transaction into a block on the PR#3 node (regtest mining stands in for real proof-of-work here — no actual mining claim, just proving consensus validity).
• The block relayed to the stock node automatically. Checked from the stock node's own side: chain tip matches, the transaction shows confirmations:1, and the 1-satoshi output is live in its own UTXO set — permanently, on a node that would never have accepted this transaction pre-confirmation.

That's the actual mechanism: relay policy only ever controlled propagation, never validity. A miner willing to include these transactions doesn't need the rest of the network's policy to change at all.

---

Scope note on both: local, deterministic, no real money, no public network. These demonstrate the mechanism works — they're not a mainnet or signet claim, and adoption/hashrate is a separate, much bigger question from whether the mechanism itself is sound.

[Mehdalon]

3. Tried it on real testnet too

Local regtest proves the mechanism works deterministically, but doesn't touch real network conditions. Went and tried that too — mixed, honestly-reported result: testnet-exploration.md

Short version: real peer discovery and faucet funding worked cleanly. One genuine finding worth knowing about the PR's current scope — addconnection ... "dog" is hard-gated to regtest-only in src/rpc/net.cpp (predates this PR, untouched by it), so on testnet today a dog connection can only ever form organically between independently-adopting peers, not forced the way the regtest demos and the test suite do it. Two things didnt pan out, reported as-is rather than worked around: real testnet3 IBD would take roughly a month and a half at observed sync speed (not attempted further), and a real stratum-pool CPU mining attempt (226M hashes over 900s against a live pool) found zero shares, consistent with CPU hashrate being many orders of magnitude short of anything that matters.