Overview Description: When network goes down in middle of page view connection should give timeout message, not 'connection refused'. Steps to Reproduce: 1.) Launch app, go to www.netscape.com for example 2.) click on a link to load a page 3.) while load is in progress, remove network cable Actual Results: After time period I get a 'connection refused' message. Expected Results: you should get a timeout message Build Date & Platform Bug Found: NT 2000052408

->ruslan

->

Yeah - it should get the timeout message. Setting milestone to future

pulling in ruslan's necko bugs ->darin

seems like it didn't change the last time...

mass move, v2. qa to me.

*** Bug 129502 has been marked as a duplicate of this bug. ***

I'm curious - is anyone working on this fix ? It still shows in 2002101612 and can be damn annoying. It does it with mailnews as well as the broswer, so it's a basic networking problem.. whenever it gets a timeout, it shows the user the "connection refused" error box, which is completely incorrect behaviour. A connection refused is a TCP RST packet being sent in response to a TCP SYN. It means that 'the server was contacted, and replied, closing (okay, it never fully opened, so maybe that's a bad word) the connection' This is entirely different from 'the server did not respond', which is what actually happened. It's very reproducible.. pull out the uplink cable to the network, set up a broken forced route to an address, or do anything else to cause the target IP to not respond. This behaviour may be a result of an incorrect socket message being passed back from windows - I have noticed that windows is especially bad at passing conn_refused back to the application layer (or many programmers are good at ignoring it). Some possibly related info: In Netscape/Linux , if you attempt to connect to a closed port on a responding IP, you can see in a packet trace the SYN and RST packets, which then cause an immediate user error. Under windows, watching multiple applications (telnet.exe, netscape.exe, iexplore.exe) I have watched applications blindly ignore SYN/RST packets, and wait for maybe a minute to see if the connection opens.. even though the packet trace clearly shows what happened.. so from that, I think that windows never correctly reports the connection refused to the app layer. Maybe the connection refused message in mozilla was a kluge to try and display the correct message in windows if it failed to connect, but went too far and displays it for timeouts as well ?

Based on the number of filed bugs, we have many more problems with error handling for connection establishment failures, so I think futuring of this problem is reasonable. This also is a problem that needs to be tested on a per protocol basis. Darin, is necko lacking some underlying plumbing, or is this another "Networking error needs to be handled by docshell" bug?

we need to add extra code to necko to fix this bug.

There should be some eventual error, although the timeout period would vary by operating system.

-> default owner

UI around connection lost has changed a lot. WFM? FF now displays ... Unable to connect Firefox can't establish a connection to the server at google.com.

Resolved per whiteboard