Comment 12 for bug 1251394

Thanks for clarifying, Dan. I can't really speak to the ingest internals, at least not without doing some research first. It seems likely there's room for optimization, but I'm not sure.

Is it fair to say that if this feature produced a net zero change in ingest speed, that your concerns would be addressed? Or, do you see this feature as something that must be addressed as a separate process from ingest so that it can run significantly faster and, if so, why exactly? (The obvious benefit would be quicker propagation of display field modifications, but I don't know if that justifies using a separate structure).

I'm curious, Dan, in your 1k records / second test, were you writing output to database rows or to STDOUT?