I can go that route, but it would be useful to have this as a basic implementation in the meantime. In that sense, having an interface that will persist may be more important than the details.
Can you say more about why this wouldn't work on a given runner? I'm also not clear that the unbounded behavior is an issue. I don't have experience with streaming pipelines, but it seems like the two reasonable possible behaviors would be to wait for the entire to_wait_for (i.e. it would have to be bounded), or else to wait just within the window. Could either, or both, of those behaviors work with this implementation? [ Full content available at: https://github.com/apache/beam/pull/6343 ] This message was relayed via gitbox.apache.org for [email protected]
