Ramon

@basvanbeek I don't think the idea is terrible. Also, let's avoid that kind of discouraging language please. Having a error key with a string payload may work fine for a...

Should this issue be moved to the opencensus-java?

(Copying my comment from #50) IMO the right thing to do is to not permit invalid view name / exporter combinations. We could either throw an error or just log...

I guess I'm struggling to see why we should bother with it at all. The view name string is totally under control of the application developer who registers the exporter...

@jkschneider I don't really see views as instrumentation, more like configuration. Internally at Google they are defined in text protobuf files. We've been talking about having a similar configuration system...

@jkschneider we could certainly do a better job at documenting these concepts in the abstract. Here's the relevant code: https://github.com/census-instrumentation/opencensus-java/blob/7106a4737b44f4e55be6c4e397a4cf01e64ba220/api/src/main/java/io/opencensus/stats/View.java https://github.com/census-instrumentation/opencensus-go/tree/master/stats/view It wasn't my intention to shut down the conversation....

I think there's a major difference between Micrometer and OpenCensus that is material to this discussion. In Micrometer, Counters, Guages, etc. (Meter implementors) are given names that are ultimately what...

My opinion is to not retry. All data is best-effort anyway and retrying adds complexity and the possibility of cascading failure.

@rakyll > Subscription can be done per exporter. Was this how it worked before? I really like this but seems like it doesn't work like that any more.

This only works on the Span in the current process. If you really want to get the whole trace you need need some way to tell your caller to start...