SNIP>>>Do we FLUSH() or do we do the Win32 FlushFileBuffers API Call?
>>
>>Frankly, Terry, that MSKB article confuses me... it directly says things happen differently than some SDK pages I read recently. Maybe because it is describing Win NT?
>>
>>Now I'd also say that we learned recently that the FLUSH command relies on the OS' "lazy writer" (can't see what else it would be) to accomplish it's work (MSKB #281281), so I can't see how the FLUSH command would ameliorate anything here.
>>
>>Note, also, that the MSKB article Paul supplied speaks to Win NT specifically.
>>
>>so... I don't know for this instance.
>
>Is it for "TableUpdate" only or would it also be required for "REPLACE"?

Well, in the context outside of the problem described by this thread and more to the MSKB article Paul referred to, it is when you know you WANT to WRITE.
So it would be after a REPLACE if there is no buffering in effect or it would be after TABLEUPDATE() in cases with buffering in effect and it would be after an END TRANSACTION (I suppose) when applicable.
But we all also know that there remains a problem with VFP or the OS or ??? when it comes to TRANSACTIONS (ref MSKB #293638).

The confusion caused to me by the Paul-suppled KB article is that it says that caching is done locally by the 'redirector' and THEN data is passed to the (file) server later. Remember, this is describing NT.
The stuff I recently read, pertinent to Win2K and later I think, says that I/O on a client is intercepted by the 'redirector', passed to the (file) sever, acted upon there, then the server cache may be updated and the data is passed back to the client 'redirector', then the local cache may be update.

cheers