DanielLeens commented on issue #11790: URL: https://github.com/apache/seatunnel/issues/11790#issuecomment-5307588462
Thanks for opening STIP-36 and for keeping the contract boundary separate from #11402. I re-checked the current `dev` source at a high level before replying here. SeaTunnel already has the raw schema-change building blocks in the CDC and sink path, but it still does not provide one job-scoped timeline / decision record that explains how one schema event was routed, filtered, applied, ignored, or failed across multiple targets. So the gap this STIP is trying to close is real. The proposed v1 boundary looks reasonable if it stays explicit about three things: 1. this is an additive observability/read-model contract, not a new correctness state machine; 2. ordering / replay / restore semantics remain owned by #11402; 3. the stable public record should contain framework-owned facts first, while raw DDL / connector-native payloads stay out of the default response or remain clearly optional/redacted. The review points I would keep sharp in the draft are: - how `schemaChangeId` is created and propagated end to end; - how one source event maps to multiple target outcomes without hiding partial success/failure; - what bounded retention / truncation contract v1 guarantees for finished jobs; - how recorder failure is reported without changing the underlying schema-change result. If the draft keeps that boundary, this looks like a good STIP to continue with a PR. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
