Github user davisp commented on the issue:
That seems fairly convincingly just all around better than the default
implementation. The only real worry I had was RAM usage with the vectored reads
but that appears to be a non issue (and even an improvement).
If you want to try the tests that @theburge suggested that'd be fine but
I'd also be willing to start making final tweaks to get this merged.
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket