In article <Pine.LNX.4.44.0302130135150.17561-100000@trillian.beachbash.net> you write:
>On Wed, 12 Feb 2003, Ben Harris wrote:
>
>> In article <Pine.LNX.4.44.0302102204540.1771-100000@trillian.beachbash.net> you write:
>> >This is the built in ethernet interface on an 8500.
>...
>>
>> Yes.  <http://www.NetBSD.org/cgi-bin/query-pr-single.pl?number=12088>
>> includes a couple of fixes.
>>
>
>Okay cool, I'll have to try those fixes then. (though they're older than
>1.6 arn't they?)

They are indeed, but the PR hasn't been closed yet, so I assume they're
still relevant.  I haven't seen that error since upgrading to 1.6, though,
so I could be wrong.

>Oddly, I'm also now getting a different kind of network error:
>daniell@tnt:~$ talk daniell
>talk is aliased to ytalk (multi-user talk) ...
>Received disconnect from 64.81.199.16: 2: Corrupted MAC on input.
>[DL-hyperbolic:~] daniell%
>
>The machine's mc0 doesn't go down during this though, and there's no
>mention of the event in the messages log.
>This kind of rude disconnect happens randomly.  I'm supposing it has
>something to do with the machine being a certain MAC at 192.168.x.y
>inside the network handled by a netgear router that's doing DMZ for that
>ip to the above IP with a different MAC.  hm if that's the case I wonder
>if forcing the "WAN" port's MAC to be the same as the machine might help.
>No, actually I doubt it would.

Erm, the "MAC" in that error is a "message authentication code", and refers
to part of the SSH protocol.  It has nothing to do with the "medium access
controller" in your Ethernet interface.  Since SSH runs over TCP (which has
checksums), a corrupt MAC probably means that either one of your SSH
implementations is buggy or something is going wrong with your packets
before the TCP checksum is put on.

-- 
Ben Harris