>Mike,
>
>I see 2 possible points to make such code less smelly.
>
>First is if a specific order has to be present for further processing *as well* as a filter.
>If the filter blocks most (read more than 85%) of the data, handling after creating the temporary
>idx can be much smoother.
>
>Second is if (heavy) editing on the columns in the filter is to be expected: using a temporary filtered index
>gives sometimes better performance in this case than using the filter.
>
>As there might be similar points we can't thiink of at the moment, this would be a case for me to call up the previous dev, even if he was working for another company by now (unless there was some hint in the comments or [s]he has left mud puddles all over the place <bg>)

Well, considering the set filter so totally outperformed the dynamic index, it's not really an issue. I guess the point I'm coming to see is NOT everything should be part of an application. There have to be some worst practices if there are also best practices.