>I learned it back in the batch days from some guys at IBM.
>Back then - before debugging was around - we'd do a binary chop of the data to see which record was exposing the issue. It was tedious but it never failed.
>With debugging, it's rarely needed.

I had one recurring situation where I'd need it - an overly long sql select with too many fields. Since SQL server isn't too informative about which field is the culprit, well, with one field per line, comment out half of the fields and see when the error goes away.