>On Sun, 4 Feb 1996 John Henders <jhenders@bogon.com> wrote:
>
>> In <9602021309.AA0027@noss.zetnet.co.uk> pm@zetnet.co.uk writes:
>>
>> >The new (1.3.5x) ARP code sends a unicast "confirm this ARP entry" request
>> >to all resolved hosts in its table every 5 or so minutes. If the host
>> >fails to respond to a certain number of them, its entry gets a 0x0 flag,
>> >meaning unreachable. This entry in the table is NEVER timed out, and
>> >subsequent ARP requests are never made, and the host remains unreachable
>> >until manually purged from the table.
>>
>> Is this the right thing to do? I know it will definately cause problems
>> with Livingston Portmasters, which occasionally seem to stop arping dial
>> in connections for a while. If this behaviour is correct, it means
>> having to hand delete the host unreachable arp table entry before that
>> port will route again.
>
>Hmmm. Does this mean that any application that is using that net
>route will be terminated until *manually* fixed?
>
>Is this a general thing in 1.3.x?
>
I'm sorry, I found this thread only today.
It is not bug, and it is not feature, it is normal behaviour.
ARP module will try to get HW address of this host
until you will try to send something there.
Do not send anything and this entry will be expired in 30sec!
If the host that cannot answer to ARP requests but sends some
packets to you (f.e. pings you or tries to telnet to you),
this entry will show up in ARP table until you turn off
that silly host.
Alexey Kuznetsov.