Walter,

The only reference I can think of was personal experience with a retail chain a few years back where each store had a 3 station peer-to-peer network, 70 different stores, and between 5 and 10 were corrupting or duplicating every night.

We wrote routine after routine to try to find out what was happening. What we ended up finding out is that VFP would FLOCK() the table, grab the number, update the number and UNLOCK fine.....as far as VFP was concerned. The OS would, sometimes, not do the physical write right away and might queue it up behind the same request from another workstation, thus "spranging" the keys. See? We were able to reduce the problem by playing with the Win buffering on the machine the tables resided on, but never completely eliminated it until we said "to hell with this" and converted everything to views and SQL Anywhere.

Since that episode, which consumed my life for about 6 weeks, I no longer trust this method of maintaining keys in a heavy volume situation. It's just not worth it to me when I can very easily use a GUID. Or even SYS(2015) in some cases.

There used to be a Microsoft KB article about this issue but it was not Fox specific and just talked about how buffering and locking could be a problem on peer-to-peer Win95 lans.


>Do you have any references to the described problems ??