An RPOW server accepted a valid proof or earlier RPOW, marked it spent and issued a fresh equal-value token. An IBM 4758 secure coprocessor and remote attestation constrained the server, but ownership still depended on one online registry rather than decentralized consensus.
Finney turned Hashcash from an application-specific anti-abuse stamp into an input for digital collectibles. Creating the first token required SHA-1 proof of work that was expensive to find and cheap to verify. [Hal Finney — RPOW overview] [Hal Finney — RPOW theory] [Hal Finney — RPOW launch announcement]
Reuse did not mean presenting the same bit string repeatedly. Each owner spent a token once to the server and received a newly signed replacement that the next owner could exchange again. [Hal Finney — RPOW overview] [Hal Finney — RPOW theory] [Hal Finney — RPOW launch announcement]
The client submitted either fresh Hashcash or an RPOW plus a blinded output request. After validation the server consumed the input and returned an RSA-signed token without needing to link the new token to the old one. [Hal Finney — RPOW theory] [Hal Finney — RPOW launch announcement]
The server created only an RPOW of the same work value as the consumed input. Separate denominations represented different Hashcash difficulty levels, aiming to prevent the operator from inflating claims on computational work. [Hal Finney — RPOW theory] [Hal Finney — RPOW security model]
A central spent-token database rejected every previously exchanged proof. This solved double spending operationally, but server reachability and correct state remained essential to every transfer. [Hal Finney — RPOW theory] [Nakamoto Institute — archived RPOW source]
Finney ran the server inside an IBM 4758 Secure Cryptographic Coprocessor. Its tamper resistance protected signing keys and persistent state even from the physical machine’s owner. [Hal Finney — RPOW security model] [Nakamoto Institute — archived RPOW source]
The coprocessor could sign an attestation of the software configuration it ran. Clients compared that measurement with the hash of published source they trusted before sending tokens. [Hal Finney — RPOW security model]
RPOW replaced ordinary institutional trust with a narrower combination of IBM hardware, manufacturing keys, audited code and one server state. Attestation reduced operator discretion; it did not create permissionless consensus. [Hal Finney — RPOW security model] [Nakamoto Institute — archived RPOW source]
The service was explicitly experimental, personal and non-commercial, and its tokens had no promised monetary value. The original project ended, while source code and documentation survive as an archive. [Hal Finney — RPOW overview] [Nakamoto Institute — archived RPOW source]
RPOW demonstrated transferable proof-of-work value before Bitcoin. Bitcoin instead makes independent nodes validate a public chain and uses competitive mining to order spends, removing RPOW’s single attested registry. [Hal Finney — RPOW overview] [Bitcoin whitepaper]
For the clearest picture, read this entry together with Hal Finney, Hashcash, Proof of Work, Bit Gold, Double-spend, Bitcoin. The reverse links also lead from Hal Finney.