>>I am sharing a table over a peer to peer network, and am a little confused about how local memory caching affects my data. If I append a blank to a table, the machine that appended the blank will see that record, but will other machines on the network? Would they have to wait until that machines cache is written out of cache? If the file server were to be shut off would it necessarily have that record saved? Is there anyway for VFP to even know if the data was really stored versus put into a cache by the OS? Thanks for any input on this subject. I'm trying to figure out how a customer could have gotten two different record counts from the same table.
>>
>
>Caching and buffering are two different issues, but essentailly, the only machine that should physically cache data pending writes is the "server" machine - and since it's what is doing the file I/O, it should know the state of the current table no matter who has written or appended to it.
>
>Information buffered (ie using record/table buffering), where record(s) exist that have not been written to the server yet, are not visible to other users (or even other data sessions for the same user with private data session. If records are present in the buffer and someone else opens the file, new records in buffers, not yet committed by a TABLEUPDATE(), would not be visible.
>
>If someone kills the server without all records being written to disk from all stations, then a difference in record counts is the least problem you can anticipate. Killing a server with data pending in cache or in buffers at the server or client is going to at a minimum lose the data not yet committed to disk.

Thanks Ed,

No, I am not talking about buffering at all. Thanks for the caching info. Can you think of a way that RECCOUNT() Would come up 5 records short? That is really the question I'm trying to find an answer to.

Thanks,
Marcus.