NeverENG opened a new issue, #42: URL: https://github.com/apache/dubbo-net/issues/42
## What problem would this solve? DubboNet currently discovers **remote Providers for consumer calls** through ZooKeeper or Nacos. Registry selection and metadata-center setup are tied to those two backends, so an application using etcd v3 cannot configure DubboNet to discover its Providers or obtain metadata from an etcd-backed metadata center. This is a capability gap, not a regression. In particular, the existing ZooKeeper/Nacos consumer registration is **not** Provider-side service publication. dubbo-go has separate [etcd v3 registry/service discovery](https://github.com/apache/dubbo-go/tree/6b04ac04bfc868060995dc9fa4e1dcc5138ad02b/registry/etcdv3) and [etcd metadata report](https://github.com/apache/dubbo-go/blob/6b04ac04bfc868060995dc9fa4e1dcc5138ad02b/metadata/report/etcd/report.go); their existence does not by itself establish wire/key-layout compatibility with every Java Provider deployment. ## Proposed scope Track **two independently verifiable adapters**, preferably as separate implementation PRs: 1. **Consumer-side etcd v3 registry discovery:** configure an etcd registry, read the selected discovery model's Provider records, convert supported `dubbo://` and `tri://` records to endpoints, and react to updates/removals. Define the supported key layout and whether this first slice is classic interface-level discovery, application-level discovery, or both; do not silently claim the other model. Specify initial snapshot vs. watch ordering, reconnect/re-watch, empty and filtered results, and disposal behavior. 2. **etcd metadata-center adapter:** independently configure the metadata center (which may differ from the registry), read the method/type definitions needed by existing consumer invocation and cache-refresh flows, and define the actual compatible layout/version and missing/incomplete-definition behavior. Evaluate publication/reporting as a **separate follow-up** tied to a future Provider/export lifecycle; do not add write APIs solely to make the consumer adapter look complete. Integrate with the existing ownership boundaries rather than branching more etcd SDK logic inside `DubboClient`. PR #36 (Provider discovery/metadata ownership) should be accounted for when choosing the integration point; avoid a second callable-Provider cache. Preserve existing ZooKeeper/Nacos behavior and public calling APIs where possible. ## Acceptance criteria - [ ] Document the targeted etcd v3 discovery and metadata layouts, supported Dubbo versions/discovery model, configuration and auth/TLS expectations **before** claiming compatibility. Confirm against real records; do not assume ZooKeeper paths map directly to etcd keys. - [ ] Consumer discovery can obtain supported Provider endpoints and process add/update/remove events; initial-read/watch races, reconnect, cancellation/disposal and empty results are covered by focused tests. - [ ] etcd metadata reads feed existing signature/type resolution and refresh correctly, including absent, partial and invalid metadata, without taking ownership of business-call Provider selection. - [ ] Registry and metadata addresses can be configured independently; existing ZooKeeper/Nacos paths continue to pass their tests. - [ ] A reproducible etcd-backed integration fixture verifies calls to an actual compatible Provider and metadata reads. Label results as tested configurations, not blanket etcd/Dubbo compatibility; update English and Chinese docs/support matrix accordingly. ## Out of scope - Implementing a .NET Provider server, exporting services or publishing `/providers` records. - Claiming application-level discovery, metadata publication, or full parity with dubbo-go until those paths have dedicated implementations and interoperability tests. - Changing Native/Triple wire protocols as part of the registry adapter. Related: #16 (responsibility boundaries), #36 (unmerged discovery/metadata ownership refactor). -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
