Colin & Larry,
Years ago I had a Clipper app that worked like this, except I had a log file which kept a record for each change made. Record had the table, PK, fieldname, old value, new value and type (since values were stored as character data in memos) plus a few flags for New, Deleted, etc.. Only the log file needed to be uploaded, which saved a lot of time on old modems. We didn't flag the original tables, instead we printed a report of "Present value in master table doesn't equal old value OR new value, don't update" entries. We found that we only had 2-3 per week of these, although files from 10 branch offices plus the main handled about 10,000 log records per week.

Barbara


>Sounds good. I wrote an app very similar to this except it was for users to take data to a job site, so they logged into the network, downloaded needed data to notebook, updated on job site and when back in the office logged in and uploaded changes.
>
>My approach to updates was this:
>
>1. Like you, downloaded records were flagged with user id and were readonly to all other users while "checked-out".
>
>For upload:
>
>2. SET DELETED OFF
>3. SCAN / ENDSCAN through local tables
>4. For each record, SEEK in master tables
>5. If local record was deleted, delete master record, otherwise replace all fields with fields from local record
>6. Remove user id.
>
>HTH