David Schlosnagle
David Schlosnagle
OpenJDK [tracking JDK-8246039](https://bugs.openjdk.java.net/browse/JDK-8246039)
Yeah, thinking we might want to use ASCII character codes 29, 30, 31 as separators, or go full structured serDe like Jackson. Thoughts? Will also add tests for `user-agent: Mozilla/5.0...
@carterkozak for thoughts as he was looking for similar recently
Relates to #8
@carterkozak curious for your thoughts on this again. I can rebase on develop if interested.
Definitely planning to hold on this until calmer seas
Related : [JDK-8292989](https://bugs.openjdk.org/browse/JDK-8292989): Avoid dynamic memory in AsyncLogWriter [JDK-8299415](https://bugs.openjdk.org/browse/JDK-8299415): Saturate the async log buffer more often
Reopening as https://github.com/palantir/sls-packaging/pull/1452 reverted `-XX:+UseStringDeduplication` until we can more broadly performance test. G1 and Shenandoah GC support string deduplication as of JDK 17 https://openjdk.java.net/jeps/192 https://bugs.openjdk.org/browse/JDK-8264718 Only enable on JDK 17+...
Yeah, I expect that we need to allow selection of Java 8 or 9, and derive appropriate default flags for each. We'll likely need to enable the module escape valves...
Is this resolved by https://github.com/palantir/tracing-java/pull/166?