I don't think we can, or want to make this bump.

The Kubernetes Client which AFAIR the main reason we have OkHTTP in the BOM is still on 3.x and doesn't plan to move to 4.x anytime soon.

Agree that makes sense, however given that v4 makes an explicit effort to be fully backward compatible with v3, is it not worth trying? The functionality used by the K8s client (and others) would be standard functionality that should work, i.e. let the integration tests do their thing?

TBH, we are really against dragging in any sort of Kotlin related thing as a dependency.
I'm -1 but let's see what others say.

cc @iocanel @manusa @gsmet @maxandersen

Yeah the kotlin stdlib dependency is problematic unfortunately :/

At the moment the alternatives are ranked by simplest to hardest (from my current POV :)

• stay on okhttp v3
• migrate away from okhttp v3 to vertx or other non problematic http client
• find a way to convince everyone the kotlin stdlib is ok

One of the main reasons we picked okhttp as an http client in the first place, was that it didn't bring in lots of transitive dependencies. Unfortunately, this is no longer the case.

I have terrible experiences with frameworks bringing in general-purpose libraries (remember guava?) and I'd like to avoid that.
If I had the option I would even avoid using Jackson.

So, I am leaning towards using vertx.

A big +1 for Vert.x.

Hi, I have no idea about the implications of adding the Kotlin runtime to the build.

The claim from the project maintainers is that OkHttp4 is backwards compatible with OkHttp3. I'm not sure if any of the projects using Quarkus might have tried tweaking the dependency configuration in their pom.xml to make use of OkHttp4 and what the result was.

I really like the Vert.x idea (so +1). There would be a lot of work involved, + real breaking changes (especially for those who implemented custom interceptors, etc. for OkHttp), but at least we'd be depending on something we could rely on.

I created a new issue in fabric8io/kubernetes-client#2764 requesting migration to the Vert.x HTTP Client.

Closing as upgrading it may not be a good idea given the reasons stated above