On Oct 23, 2008, at 8:03 PM, Jens Alfke wrote:
On Oct 23, 2008, at 4:42 PM, Greg wrote: Very interesting. I'm still not sure though this was the best decision on (whoever's) part, because the socket becomes unreadable and unwritable while you're on that other network; essentially a zombie socket, which isn't very useful anyhow. Perhaps it will reanimate if you switch back to the other network, but why would you do that or want that behavior?
For one thing, the network change might be very short-lived. If you have to briefly unplug and replug a network cable to move a machine, or if the cable accidentally got yanked out and then put back, or if you lose your WiFi signal for a second and then regain it, it can be nice that sockets continue on instead of all suddenly disconnecting. Especially in a server environment.
Well, that implies that the situation "User initiates network switch", and the situations you gave (losing wifi, cable unplugged, etc.) are "the same", or can't be differentiated. That may well be the case in the way the OS X operating system is written, or the way TCP works, I understand, I was just pointing out that they are not, and in this context, leads to odd behavior.
- Greg |