I'm having problems with mail transfers.  When I send a message from 'pyy'
to microsoft.com, the message never gets through.  Here's what tcpdump
says:

11:33:56.707881 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: . 0:512(512) ack 1
                win 16384
11:34:52.395512 mail1.microsoft.com.smtp > pyy.jmp.fi.2571: F 1:1(0) ack 0
                win 8104 (DF)
11:34:52.396154 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: . ack 2 win 16384
11:34:58.714630 mail1.microsoft.com > pyy.jmp.fi: icmp: ip reassembly time
                exceeded
11:35:00.710560 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: . 0:512(512) ack 2
                win 16384
11:36:04.713089 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: . 0:512(512) ack 2
                win 16384
11:37:05.077396 mail1.microsoft.com > pyy.jmp.fi: icmp: ip reassembly time
                exceeded
11:37:08.715594 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: . 0:512(512) ack 2
                win 16384
11:38:12.718261 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: . 0:512(512) ack 2
                win 16384
11:39:16.720203 pyy.jmp.fi.2571 > mail1.microsoft.com.smtp: R 581:581(0) ack 2
                win 16384

etc.  My network looks like this:

'pyy'---ethernet---'gw1'---PPP---'gw2'---ethernet---'ISP router'

I can see the same packets on 'gw2' with tcpdump (only the large packets get
fragmented during PPP transfer).  What does the "ip reassembly time exceeded"
message mean - is this timeout occurring at the remote end?  Does the network
drop one of the IP fragments or what's going on?  How could I track down the
problem?  I have been seeing this for a long time, but first I thoughtit was
the NetBSD/Amiga system I was using as a router.. but I still have the same
problem after removing that machine.  I also have problems _receiving_ mail
from one machine in the world, but I haven't checked that with tcpdump yet.

Any ideas?

  -jm


-- 

                     ---> http://www.jmp.fi/~jmarin/ <---