>>I don't understand...
>>How does the file server know to put all other users on hold while waiting for the WS to send the write, or for the network connection to pass it on?
>>Is there a lock on the file header? Then it shouldn't matter how slow the WS is to send the write, because other users won't be able to initiate another append.
>>If there is no lock, then it doesn't matter how fast the WS is, there is always a chance to get duplicate keys, not just once in while.
>
>With a shared file, you only get to write when you have an appropriate lock. But consider what happens if a workstation crashes while it holds a lock.
>
>During normal use, the server and workstation periodically send "heartbeat" signals to each other to ensure the connection is still valid. If a workstation crashes and no longer has a heartbeat, eventually the server is going to kill the connection and all of its associated locks. At that point a VFP table may be in an inconsistent state, due to no fault of the server.
>
>So, it *does* matter how long it takes for a workstation to complete a write. I don't know the details on how long that is or under what conditions locks may be broken by the server OS.

I feel the need to sit down with you and use pencil and paper to clarify this.
The dialog reminds me of the Romanian joke about technical universities students: Q. In what ways are an Engineer and a dog alike? A. Both have an intelligent look, but they can't express what they think. Talking about me, of course.
Maybe we'll talk about it at the next VFUG meeting.