I agree with Enrico that it's better to have a config option. Also, we cannot simply replace the PulsarVersion call with the DynamicPulsarVersion call because the client version string is now constructed as:
String.format("Pulsar-Java-v%s", PulsarVersion.getVersion()) It's a config of client version string, not pulsar version. Moreover, in your proposal, you mention the case of client c++ at first, but don't talk about it later. Is the scope of this proposal in the Java client only? Best, tison. Enrico Olivelli <eolive...@gmail.com> 于2023年3月4日周六 06:38写道: > Yunze, > > Il Ven 3 Mar 2023, 12:31 Yunze Xu <y...@streamnative.io.invalid> ha > scritto: > > > Hi all, > > > > Based on the previous discussion [1], I created a proposal to support > > configuring client version at SDK level: > > https://github.com/apache/pulsar/issues/19705 > > > > I've added more explanations in the motivation part, let's use this > > PIP as a subsequent discussion of [1]. > > > > BTW, there is a PR [2] in the pulsar-client-cpp repo because the > > motivation is more meaningful for the C++ client. > > > > I understand well this problem, we have it for the cited clients but I also > see the same issue for other libraries based on the Java client, like the > official Apache Pulsar Reactive client. > > I also see this problem in Startlight for JMS that is a JMS client for > Pulsar that is based on the Java client. > > While I agree on the problem and on the solution I think that a static > field is not enough, we have some problems: > > 1) there may be multiple usages of the Java client in the same JVM, and you > want each client to report correctly its version > > 2) we would need to use the Java security Manager in order to prevent > malicious code to modify the version or some other mechanism to prevent > overriding the version. > > I believe that in the case of the Java client is is easier to add a > configuration entry to the Pulsar Client Configuration. That would become a > field in the JavaClient. So each instance can declare its version and also > malicious code won't be able ti easily tweak the version (because it won't > be a simple static method call) > > Enrico > > > > > [1] https://lists.apache.org/thread/n59k537fhthjnzkfxtc2p4zk4l0cv3mp > > [2] https://github.com/apache/pulsar-client-cpp/pull/208 > > > > Thanks! > > >