I've added a new 233MHz machine with W98 into a network yesterday. Here's the layout: two P166 machines, one acting as a server, both under W95. Two 486 under dos6.2, acting as workstations. UTP cables, little 8-port hub. Everything worked fine so far.

Now we added this machine and a HP6P laser with it as a local printer. The apps running are FPD2.6, planning to add some VFP soon. Everything worked fine to some extent, when it began closing files on me. I've seen it opening a file without an index, or closing the .fpt, or just stopping the indexing in the middle on the new machine only. The DOS message was "The device does not exist on the network reading drive F" (including the odd amount of spaces). Even more funny thing was that the files still seemed to remain open on the server side; reopening them worked just fine most of the time, but sometimes we got "attempting to lock... please wait", and once we even got a marquee (scrolling text) in the Wait Window with apparently Replicate(chr(6),3)+" waiting for..." or something - I forgot the exact text, I was just shocked with FPD displaying a marquee on me for the first time after so many years. I never imagined it's capable of doing it on its own.

Then we thought it's something wrong with its DOS task settings, and tried mass copying from one machine to another. Everything worked fine and smooth, except when this W98 machine was reading from this W95 server. It complained of losing the mapping to F: drive, but when we clicked on it in IE4, we could start copying all over. We tried copying from the other W95 server and it worked fine. (We actually moved an 80M directory around, and it was done below 3 or 4 minutes each time). We increased the PCI latency time from 48 or so to 128, but it just delayed the moment of losing it.

I've reproduced the behavior in the office today - same layout (w98 workstation with w95 server, UTP with 8-port hub). While monitoring in NetWatcher, I've noticed the other machine completely disappeared from the list of logged-in users. First retry brought it back. Then I've run the indexing routine in a loop, and it ran fine 150 times, then suffocated on 153, 213, ... pretty random.

Any ideas on where to look for culprit (or a scapegoat, never mind)? The next thing I have to tried is to move everything to the new machine and use that one as a server, though it's the solution I don't particularly like - it's the boss's machine, and bosses are probably not the right people to sit at the server. It worked fine, though.