emccall@eagle.wesleyan.edu (03/22/91)
I'm adding support for routing IPX to Vance Morrison's PCroute. Things are working pretty well right now, except: workstations are logged out of servers after 5 minutes of inactivity when logged in through PCRoute. Using our Network General Sniffer, I found that the server was sending keep alive packets both to PCRoute and to the workstation logged on through PCRoute, but with the wrong network number. Eg, in the network: 12 <PCROUTE> 101 +-----------| |----------+ <server> <workstation> packets would be sent to PCROUTE on net 12 for connection number 16, and to workstation on net 12 for connection number 19 instead of a single packet sent to workstation on net 101 for connection 17 (which is the number I saw in the Create connection reply). Two more things: When logged in to a server through a cisco router *and* PCRoute, things work okay (keep alive packets look like they should). Eg: 100 <cisco> 12 <PCROUTE> 101 +--------| |-----------| |----------+ <server> <workstation> Also, in order to test the lower levels of the code, and to save time, I didn't do a full SAP implemetation. Instead, SAP broadcasts are bridged through the router to every interface (except the one the SAP packet was received on). PCroute must doctor the packets, changing network number and intervening network count as needed. (I'm actually regretting that decision -- I now think this was may have been more trouble that doing it correctly would have been...) Bridging these packets and routing the rest is suspect, but everything else seems to work okay. I'm wondering where the server (SPX, I guess) gets it's information on where to send the keep alive packet to. It doesn't use bindery information, because USERLIST /A shows the correct information, but I certainly can't find any literature on it. If anyone thinks of where to look for this information, or knows of anything else I should try, please let me know via e-mail. Anyway, thanks for your time, and any help that can be provided.