Upgraded from a Novell server to a windows 2003 server.  System seemed to be working fine, but after users were in for a while, get errmsg.b, then when they try to go back in get error 3105. It looks like I moved my dba over the hard way after looking in the forum. Here is what I did:
 
1. I installed pervasive 2000i and dba classic to the 2003 server with the dba 2004.1 update disk 2-14-2004 (the last update provided by dba before they stopped support).  Then I installed the iss tech 08-11-06 and 10-31-06 updates. (These were both installed on the Novell server). Then I created each company with identical company codes as the companies on the novell server. I unloaded the pervasive engines on both the novell server and the windows 2003 server. I deleted the files from the newly created company directories on the new server ie dbamfg\pl,dbamfg\ds etc and the dbamfg\default directory.  Then I copied the files from the corresponding directories on the novell server and pasted them into the directories on the new server except for the loc files (~pvsw~.loc). I copied all the *.rtm files from \dbamfg over too. I started the pervasive services, and ran dba on the server. I logged in to all my companies, all my user setup and security was correct, I could access all the data fine etc. Created .ddf files for each company.  At the workstations, dba ran from a drive mapped as f: so it ran from f:\dbamfg, which is the same way it is mapped to the new server. In pervasive control panel on the workstations, under configuration, client & client16bit, changed supported protocols to just TCP/IP, and registered the new engine(removed the old novell one if it existed). (First off I disabled client service for netware, nwlink netbios, nwlink ipx/spx-the netware server is no longer in use.) Logged users in from each workstation and it worked fine - but it was end of the day. Today they were in till 2:00 when we started getting the above errors. Getting everyone out, stopping the pervasive services and restarting them lets everyone back in for a few hours before it happens again. A significant number of the workstations are on static ips. DHCP is controlled by the router. On the clients in PCC under client, access, Target Engine:try server, then workstation, use local microkernel enginen.  I would appreciate  any help. We do have a support contract with is tech, but Bill is gone this week, and Lynn wasn't available when I called. The 2003 server is configured as a workgroup server, it is not a domain controller.

Non of the clients are pervasive 8. They are all 2000i.

We did not reinstall the clients, just removed the old registered engine if it existed and registered the new one.  I did check, our pervasive is installed to f:\pvsw, not under the dbamfg directory.  In the f:\pvsw\clients\win directory the data1cab is dated March 22, 2001.  We never install the clients from anywhere but here (was the same on the novell server where they were originally installed from).  When you start dba the pervasive flash screen starts and shows pervasive 2000i - would a version 8 client still show this?  When we first started having the problem, before I posted this, since my documentation for error 3105 says no transport portocol is common, we checked on the server, and  it had all three communication protocols: Microsoft SPXII, Microsoft TCP/IP, Microsoft NETBIOS, so we changed each workstation to have all three also instead of just TCP/IP. I know that really shouldn't make a difference, just having the one protocol should be quicker to connect, but so far today we haven't had any errors. I'm sure just one or two workstations could be a problem, and its possible they aren't accessing anything today to cause a problem.

Just wanted to to update this for people who are using this info as reference - Lynn was right - we had an employee that set up an additional workstation and it was version 8. We checked every station and that was the only one, but when it was in use - sure caused a lot of problems.