[quote user="ccomley"]
Can I stick my oar in here as we've been trying to resolve this problem for Johnny and getting nowhere.
It's not MTU or similar. The problem exists even if the Pegasus workstation is on the same LAN as the server. It's not time-out either. The session dies off, no amount of waiting for it to come back helps.
The server is Microsoft Exchange Server 2003 (SBS2003 version) with SP2, being accessed in this case via POP3. (MS's standard POP3 server module, no special add-ins.)
The problem seems to be a specific mail item in the queue. The "manual" resolution is to use OutlookWebAccess to open up the mailbox, and move the offending item to a subfolder, or delete it, then PMail is able to fetch the other items in the queue entirely normally.
We do not think the "rogue" email is corrupt. The problem recurrs whenever that sender sends a similar missive. They tend to send mail with complex HTML content - but so far as I can work out *what* is contained within the message body shoujld just be regarded as "payload" by the POP process so it really shouldn't matter, even if the resulting email is one which Pmail can't display properly, etc. it should still download it.
I am investigating to see if it's a known issue with ES2003(POP).
I believe the *server* has a Nvidia chipset, I'll investigate changing this setting on the server and see if that makes any difference!! <later> It's a Nvidia board with an Nvidia Gig nic but there isn't a parameter called "checksum offload".
[/quote]
Agreed, POP3 process should only get hung if there was some highbit character stream in the body that could block the TCP/IP process. You might need to use "Selective mail download" on this specific message type while you have turned on the session logging.
FWIW, the server itself can cause the packet fragmentation if there is a firewall running blocking outbound packets. However is the packet discovery is turned off on the client side then this becomes a non-problem.
You might also try turning on blocking sockets using the -Z 1024 commandline option to see if this helps.
1024 Use blocking sockets; may be needed for some
defective WINSOCK implementations.