>You are right except that if you USE DBF('mycursor') AGAIN on a cursor that VFP has emulated by just setting a filter on the underlying table, then you are using the underlying table again, and any attempt at indexing will be denied. The trick, as I found was to use the NOFILTER clause to force the creation of a new cursor from the query. Thanks for your input.
>
>>Erik,
>>
>>It seems like the old USE DBF('mytable') AGAIN... is exactly what you need. Just because you make the cursor read/write doesn't mean you have to make any changes to it. Any changes that are made won't be applied to the base table unless you use CURSORSETPROP() to tell VFP to send updates, etc.
>>

An older trick is to
USE DBF('mytable') AGAIN...
index on whatever tag whatevername OF (tmpcdxname)

...where tmpcdxname is any random filename, I usually compose it using _work+sys(3), and _work is usually "C:\scratch\". By using a separate .cdx file, we don't try to create a structural index, and it doesn't have to touch the 29th (or was it 28th) byte of the table header.