>Someone had mentioned that there was an article in
>Foxtalk about using DELETED() as an index. Does anyone
>remember the issue or the actual specifics (besides it being
>a bad idea?)

Ed,

I don't know about FoxTalk, but ADVISOR May 99's answer was "It depends."

In the creation of local views, Rushmore brings the index to the local machine to process. If there are only a couple of deleted records it would bring a large index to the local machine. If there were 50% deleted records it would bring 50% of the (large) index to the local machine. If there is no index on deleted() it brings the other indexes, picks through them, and then checks the result individually for deleted() status. This may mean checking only the 100 (or whatever) records that resulted. The tradeoff is between the time to bring the large index over the network and the time to sort through the result set individually.

This of course depends on whether you have a fast/non-busy network and a slow local machine, or a slow/busy network and a fast local machine.

It also applies to any index that only has a few possible choices such as items from a pick-list.

When a second query is run the indexes are already on the local machine and can be used again, which makes the second run faster, so whether you are doing something once or repeatedly affects your approach to this problem.

It's a good article. "It depends."