>I'm offering a theory here that would account for lost data I posted about this morning. There are 3 tables on the server. A user opens them up EXCL in an app, does an hour of edits, the machine crashes, and all the edits are lost. This is with table buffering, and in spite of constant Tableupdates during the editing. All tables have edit timestamps, and no stamps were made for the editing period. All tables are fine otherwise, the editing was re-done without trouble.
>
>The theory is that somehow the vfp app copied the entire tables (which are quite small) into local memory, and performed all the Tableupdates on the local copies. The the machine crashed, and the copies were never returned to the server.
>
>I realize this sounds odd, and I've never had anything like it before, but I can't account for the data loss in any other way.
>
>I would like to think this can't be true, but can anyone tell me that this theory is not possible?

I could buy that theory only in that the tables were opened exclusively. I say this because if the tables were opened shared and this were true, other users would view old values if they opened these tables after you had issued a tableupdate. What a mess that would be. If this theory is true, this method of internal data handling sucks. We should not have to issue a FLUSH to assure writes to disk when a TableUpdate should suffice.