Hi everyone,

Following on from Ryan's thread in April [1], where it was agreed to keep
the pre-built Kafka Connect connector out of releases until its
LICENSE/NOTICE issues were fixed, and Kevin's suggestion in May [2] to
start from runtime-deps.txt, I've opened a PR that I believe fixes them:
https://github.com/apache/iceberg/pull/18019

It finishes the work Ryan started in #16147 and:

- regenerates runtime-deps.txt, guarded by checkRuntimeDeps, the same way
as the cloud bundles and Spark/Flink runtimes
- has the runtime declare the same cloud SDK dependencies as the
aws/gcp/azure bundles, so its LICENSE and NOTICE can be kept in sync with
the bundles' files
- reconciles LICENSE and NOTICE against every jar in the built zip,
including shaded code

Two things to flag for awareness/comment:

1. It removes the Hive distribution. It could come back later with its own
runtime-deps.txt if people still want it.
2. The distribution grows a little (third-party jars 243 → 257, zip 184 →
192 MB), because the bundles also declare the SDK modules behind Iceberg's
optional cloud features (S3 Access Grants, the S3 analytics accelerator,
CloudWatch metrics publishing, Google Cloud KMS, GCS analytics, Azure Key
Vault), which the previous hand-picked list left out. The Hadoop tree is
unchanged.

There are full details in the PR. Reviews welcome, particularly on the
licensing judgement calls listed under "Notes for reviewers" in the PR; I'm
not a licensing expert and would rather those were checked than taken on
trust.

Full disclosure: Claude helped me a lot with this :) I drove and checked
the work. `check` passes, including the Kafka Connect integration tests
against the installed distribution, and every jar in the zip was opened and
compared against LICENSE.

This isn't aimed at 1.12. Once it lands, I'd like to pick #15212 back up so
the connector can ship as part of a future release.

thanks, Robin.

[1] https://lists.apache.org/thread/3p7nrxfv3p1hclfww4hjdtbmfwklgok7
[2] https://lists.apache.org/thread/rvp2wh1m91h5lvz5zvxqp96f7pdxrzqr

Reply via email to