Are jvm.class.loaded and jvm.class.unloaded names correct even if .count is adopted for (monotonic) Counters?

[trask]

I believe this is the last question we have about JVM metric naming.

I think that these names are correct even if .count is adopted for (monotonic) Counters in #260, but opening this issue to discuss.

The reason I think these names are still correct, as opposed to renaming them to jvm.class.loaded.count and jvm.class.unloaded.count, is that the "implicit attribute namespace" of jvm.class seems more useful (e.g. maybe jvm.class.name or jvm.class.package) compared to the alternative namespace if we renamed them to end in .count.

[trask]

Discussed in Semantic Convention SIG meeting today, and consensus is that #260 will likely take a while to resolve given that it would also impact the prometheus mappings, and it's ok to go ahead and stabilize with the current metric names jvm.class.loaded and jvm.class.unloaded (and given that if .count ends up being recommended for monotonic counters, it will never be a MUST and so it's ok if JVM metrics diverge from the recommendation in this case particular case, especially given the namespacing argument above)