This message was deleted.
# general
s
This message was deleted.
f
I agree. To get to Maven Central, we need to identify all dependencies that aren't already in MC and either do the work to put them there, or cut them out, or swap them out for another that is already there.
For example, for Kafka, we use LI's fork of Kafka, which AFAIK is also not on MC, but we should be able to easily swap this for regular Apache Kafka. We would still use the LI fork internally but the dependency swap can happen inside LI instead of in OSS...
For other dependencies, the choice may be different. For example, the Tehuti metrics lib is developed by the Venice team already, so we could move it to MC at the same time as we do the work to move Venice itself to MC...
e
That’s another point in my list for a fully public release đŸ™‚ One day I will ask more about how we depend on the LI Kafka (If I swap it I see some NoSuchMethod errors)
The most important artifact to Publish is the basic HTTP thin client
because server side stuff can be bundled on a docker image or a tarball
The same for the Writers (Samza, PushJob, Pulsar)
f
We probably don't need the extra APIs that result in NoSuchMethod errors, but we need to evaluate that... if you've already done the exercise, feel free to document the issues you found in a GH issue.
Or in a PR with TODOs on the code locations that break the dependency swap. Either way works, as you wish.