>>>I'm probably wrong in this, but the TOP is issued after order by. So you are sorting all of them and then taking the top 21. This is 2 steps non optimized.
>>>
>>>Why not create a view without the order by, open it nodata, index on ship_no and then requery() as test 1.
>>>
>>>Test 2 create a view without the top, or order by. Same as above, but count the first 21 rows.
>>
>>Maybe I'm not quite following what you're saying, but I don't think this would work well for those really *large* tables.
>>
>>The purpose of this endeavor is to come up with method for navigating thru a dataset 20 records at a time. I select a starting point (hence the "ship_no >= m.ship_no" part) and then select 21 records (20 that are displayed and the 21st as a starting point for the next 21).
>>
>>I was shooting for a single method that would work well in both native Fox and SQL Server but maybe that's just too much to wish for, huh?{s}
>>
>>Thanks, Stephen.
>
>Dude, use the view! In a view you can tell it the volume of data to return. If you like 20, then so be it. This will work on local as well as remote views.
>
>DBSetProp('SUMVIEW', 'View', 'FetchSize', 20)
>
>__Stephen

Stephen,

How does this affect the number of records returned? Isn't FetchSize just a count of how many will be returned at one time? You'll still get all the records, eventually, right?