James Carlson 已写入: >Quaker Fang writes: > > >>--------- >>1. ARP >>Client -> ARP request -> broadcast -> key1 (encrypt) >>Peer -> ARP response -> unicast -> key1 (decrypt) >>Since only key1 is used, and driver has correct key1, the ARP can works >>well without SUNWcry package. >> >> > >So what happens when the ARP entry on the peer's side needs to be >refreshed? > >The peer will send an ARP request message, broadcasted using key2. >The client (which presumably cannot use key2) will then fail to >respond. > > > Agree.
>>I have verified this on TP-LINK(ADSL router), and DWL-2100AP(pure WiFi AP). >> >> > >I see. > >Just the same, what you're seeing is _not_ a dependency on DHCP. It's >at best a coincidence. > >What you're describing is an inability to receive broadcast and >multicast messages. As I said before, if you can't receive those >messages, neither IPv4 nor IPv6 will function properly, and failures >will be both hard to predict and hard to diagnose. Such a >configuration cannot be supported in any meaningful way, even if it >appears to "work" for some tasks. > > Agree, then SUNWcry is a must. -- Quaker _______________________________________________ networking-discuss mailing list [email protected]
