There are several buffering options available that provide strict buffers: * unbounded (just keep allocating more until you get an OOME) * bounded, shut down gracefully when full * unbounded, using disk instead of memory
I think this provides plenty of options. I'm not sure we want to get into providing an option that says "emit final results only (unless it turns out you need more memory than you allocated, in which case emit both intermediate and final results)". 1, that's pretty confusing. 2, The main justification for "final results mode" is that the destination is some kind of system that doesn't permit updates. If we *claim* to support final-results-only, but send updates instead, we risk causing harder-to-debug downstream failures. [ Full content available at: https://github.com/apache/kafka/pull/5567 ] This message was relayed via gitbox.apache.org for [email protected]
