I think that SSH will clear itself from the tcp/ip stack after a certain period of inactivity But I haen't looked at the sourcecode of any flavour of SSH. UNIX and hence Linux is very efficient at garbage collection so I expect timeouts on the tcp/ip stack. Of course Bill Gates never read beyond the preface of a book about UNIX and prefers 'persistant' connections/sessions rather that 'on demand'. DOS is just a suibset of the UNIX kernel. Though Microsoft has declared it no longer supports DOS on any version of NT (The NEW Technology..[sic] ) just click 'run' and enter CMD and well, well what do we see.DOS, netbios and lan manager! Martin
----- Original Message ----- From: "Mark Stewart" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Tuesday, October 08, 2002 4:10 PM Subject: RE: [expert] networking wackiness > > > > > While the connection is OK, try "arp -a" (don't know the W2K equiv.) > > > > and make sure the other host is in the table; when it fails, > > > > recheck... if the other host's MAC is still there (both ends), then > > > > the network is likely at fault. Could be a VLAN misconfiguration... > > > > > > Okay, so from my dhcp laptop I ran "arp -a" (yep, it's the same under > > > Win2K) while I had an active happy connection to mybox and it lists > > > mybox by its IP and MAC. Likewise mybox lists my laptop in its arp > > > output. Five or six minutes the ssh connection has died and arp no > > > longer lists mybox. mybox, on the other hand, still has the entry in its > > > listing which is consistent at least since it never has any trouble > > > ping'ing the laptop. > > > > > > Seems like this doesn't quite fit the pattern but is it still a sign of > > > a misconfigured VLAN or something else? > > > > If every piece of s/w stuck to the rules, everything would just work... > > :^) > > > > > > From what you say, it appears that the ARP entry in the laptop is > > expiring... when the OS gets a packet for delivery, if an IP-to-ARP entry > > is available, it uses that entry... if no entry, it should broadcast an > > ARP-request in an attempt to re-populate the entry... failure to re-ARP > > is a bug IMO... > > > > Once the end-systems have their ARP entries, they can communicate -- if > > there are routers, switches, VLANs, etc in between, these should "do the > > right thing" to not interfere with the packets; that's why switches > > without MAC-to-port mappings flood unicast packets... > > Okay, so I am understanding correctly. The switch between mybox and the > laptop is likely not flooding unicast packets in response to the ping. > > Now we'll just have to see if our LAN admin will see this as a gross affront > to their ego or whether they'll be interested in fixing the problem. If not > (and I'm not exactly holding my breath) I have a practical workaround. > > Thanks very much, > ::mark > > > ---------------------------------------------------------------------------- ---- > Want to buy your Pack or Services from MandrakeSoft? > Go to http://www.mandrakestore.com >
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com