ERC: Onchain Announcements

This thread is for discussion of a draft ERC: Onchain Announcements.

Summary. The ERC defines a convention for coupling a human-readable disclosure with the state changes it describes. An announcement is a bracket: an Announcement event opens it, zero or more self-targeted calls execute atomically inside it, and an EndAnnouncement event closes it. A caller-chosen id ties every effect in the bracket to its disclosure.

  • announce(internalCalls, id, description, uri) publishes a new announcement and dispatches its state changes in the same transaction.
  • updateAnnouncement(internalCalls, id) revises an existing announcement: it emits AnnouncementUpdated, tells consumers to re-fetch the contents at uri, and can bracket further state changes (for example, executing an action that was announced earlier) under the same id.
  • uri resolves to a JSON document with a required category, so that domain-specific schemas (for example, corporate actions) can build on it.
  • Assets can designate announcement references, identified by ERC-7930 Interoperable Addresses, so that a service provider or a home-chain feed can announce on their behalf.

Motivation. Operator-driven changes such as token rescalings, issuances, and parameter updates surface today only as raw effect logs, with no rationale and no link to an offchain disclosure. Real-world actions such as corporate actions also change over time: they are announced, corrected, confirmed, and executed. This ERC gives them one onchain identity for that whole lifecycle.

Some feedback we are looking for:

  1. Should finality of an announcement be expressed onchain, or is it enough to leave it to the content schema?
  2. Is the announcement reference mechanism sufficient for cross-issuer service providers and multichain assets?

The PR to ethereum/ERCs will be linked here once it is open.

I’d put finality onchain. The JSON behind uri can change after the event, and updateAnnouncement tells consumers to re-fetch it, so nothing onchain records what the disclosure said when the calls ran. A hash of the document in Announcement and AnnouncementUpdated, plus a final flag after which updateAnnouncement reverts for that id, lets an indexer check that the corporate action it shows is the one that executed.

If the internal calls are calls to address(this), the functions they reach see the contract itself as msg.sender. The easy way to make that work is to accept msg.sender == address(this) as authorized, and then any other path where the contract calls itself, a multicall or a callback, gets the same rights. Worth saying in the spec that announce checks the caller’s permission for each internal call. Same for id: scoped per announcer, so a reference announcing on an asset’s behalf can’t update an id the issuer opened.

From the wallet support side, an unexplained balance change is one of the more stressful tickets: a token gets rescaled or split, the number moves, and people assume they’ve been hacked. A short note a wallet could show next to that change, saying the issuer adjusted balances and why, would answer most of those questions before they reach support.