Stellar's Bell Clears Testnet, With the Mainnet Vote on 23 November

Stellar's next protocol upgrade is live on a test network and nowhere else. Validators voted Bell, formally Protocol 30, onto testnet on 3 November at 17:00 UTC, and the Stellar Development Foundation said a day later that the upgrade had completed. The mainnet vote is set for 23 November at the same hour. Until validators there approve it, nothing on the live network moves. Three changes ride on the release. The one touching the most software is a new kind of contract address.

One contract, many customers

CAP-84 lets a single smart contract hold funds for many customers at once. It attaches a 64-bit identifier to the contract address, so a business no longer needs one contract per user. The new addresses begin with the letter W. Existing muxed accounts, which tag a plain account instead of a contract, begin with M. The identifier is recorded in the transfer event while the funds reach the underlying contract.

"It's an upside-down M. We considered reusing M, but decided it was better to have a distinct type," the foundation's post said.

Who said it is less settled than what was said. The post names no speaker. Blockonomi puts the "upside-down M" phrase in the mouth of Jake Urban, a senior engineering manager at the foundation, and it is the only account read here that names anyone. Wallets and anything else that parses Stellar addresses must add the W type or reject it.

Signatures built for a later problem

CAP-87 lets a smart contract verify ML-DSA signatures, the lattice-based scheme NIST finalised in 2024 to replace signatures a quantum computer could break. Verification runs inside Stellar Core, not contract code, because it costs far more than a conventional check. crypto.news says three verification functions ship, one per ML-DSA security level. No other account read here gives a count. Ordinary transaction signatures stay on Ed25519. A native post-quantum signer for classic accounts is a later protocol change, not this one. Ripple set out a four-stage path of its own for the XRP Ledger earlier this year.

A clock change that changes nothing yet

CAP-88 moves ledger close timestamps off whole seconds and onto a millisecond clock. By itself it makes no ledger faster. What it removes is a floor: a network that records time only in seconds cannot aim at intervals shorter than a second. Protocol 28, which shipped in September under the name Adapter, cut mainnet close times from about 5.7 seconds to about 5 seconds. No account read here gives a close-time figure for Protocol 30.

The cost falls on tooling. The ledger header format changes, so indexers, explorers and analytics that decode raw ledger data without an SDK have to update their XDR definitions first. The foundation's own line is blunt: do it before the upgrade "or your decoder will stop working". Existing Soroban contracts keep running untouched, on crypto.news's reading. The old closeTime field still reports whole seconds, so anything reading it sees no difference.

What the 23 November vote still needs

Validators have to be running Stellar Core v30.0.0 before they can vote it in. That release went stable on 16 October. The foundation also shipped Protocol 30 versions of Horizon, RPC and Galexie, with stable infrastructure and SDK releases across 16 to 23 October. Curve took StableSwap to Stellar in its first non-EVM deployment this year, so the address change reaches contracts the foundation did not write.

Here the accounts part. crypto.news names specific SDK releases, JavaScript v17.3.0, Go v0.8.0 and a Soroban Rust release candidate. The foundation's own post, the primary document, lists no SDK version numbers anywhere. We could not establish where those numbers come from. The announcement date is no firmer. Blockonomi dates the release announcement to 20 October; crypto.news cites that date and a 23 October upgrade guide in one piece, and the post carries no date to check against. The foundation adds that all of these dates may still change. What the vote turns on is how many validators install v30.0.0 in the next nineteen days, and nobody has published a count.