If a Web server has more than one network address and a connection to the first address times out, kio tries another address immediately before timing out. Even if the connection to the second address succeeds, kio immediately disconnects it. Actually, I only have one case of this happening, but I assume the above general description is correct. This case is http://www.tcpdump.org. The host resolves to six addresses: host www.tcpdump.org www.tcpdump.org has address 178.77.96.193 www.tcpdump.org has address 132.213.238.6 www.tcpdump.org has address 69.4.231.52 www.tcpdump.org has IPv6 address 2607:f0d0:3001:62:1::52 www.tcpdump.org has IPv6 address 2001:4830:116e:2::6 Five addresses are live, but 2001:4830:116e:2::6 is not. Konqueror and Rekonq obstinately try to connect to this address every time I hit the browser reload button. Firefox also tries this address at first, but after about 250ms tries one of the IPv4 addresses and succeeds. With the default 20s server connect timeout, the following happens (Time is relative to the first packet; Dir(ection) is -> for outbound and <- for inbound): Time Dir Host TCP flags 0 -> 2001:4830:116e:2::6 SYN 0.9 -> 2001:4830:116e:2::6 SYN 3 -> 2001:4830:116e:2::6 SYN 7 -> 2001:4830:116e:2::6 SYN 15 -> 2001:4830:116e:2::6 SYN 19.7 -> 2607:f0d0:3001:62:1::52 SYN 19.8 <- 2607:f0d0:3001:62:1::52 SYN,ACK 19.8 -> 2607:f0d0:3001:62:1::52 RST With the timeout set to 30s, the pattern is the same. The first 15s are identical, then: Time Dir Host TCP flags 29.7 -> 2607:f0d0:3001:62:1::52 SYN 29.8 <- 2607:f0d0:3001:62:1::52 SYN,ACK 29.8 -> 2607:f0d0:3001:62:1::52 RST I think if the connection to another address succeeds, kio should use it, especially when it succeeds (just) within the timeout setting! With the timeout set to 120s, everything works after 30s, and the browser successfully displays the content: Time Dir Host TCP flags [0-15 same] 30.0 -> 2607:f0d0:3001:62:1::52 SYN 30.1 <- 2607:f0d0:3001:62:1::52 SYN,ACK 30.1 -> 2607:f0d0:3001:62:1::52 ACK Possibly related to bug 63088?
*** This bug has been marked as a duplicate of bug 63088 ***