https://bz.apache.org/bugzilla/show_bug.cgi?id=65311
--- Comment #4 from Remy Maucherat <r...@apache.org> --- (In reply to Mark Thomas from comment #3) > I'd have no objections if you wanted to back-port it now. I did the backport it works fine., but I'm having second thoughts. I think this might cause a performance degradation on Windows. The JDK 17 changelog with its NIO update for Windows clearly hints the selector is limited compared to Unix until Java 17 is out. In that case, having a dedicated poller for blocking IO could be better (two selectors being better than one if the selector scalability degrades with the socket count, and that's not counting the hack configs to have more more more more). Or we can say anyone really affected can try NIO2. What's the final call ? -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org