One item that was touched on briefly was the downstream reuse of the data for "other" applications. In this case, it is important for a developer to not only enforce constraints for the initial data but also to realize how the data could be used in a data warehouse environment. This becomes even more relevent for those that are dealing with "spatial" datasets. For example; in a recent application and data cleansing assignment that I was working on, spatial data was being entered by end users who had no knowledge or capability to check to see if the data they were entering was actually referencing a valid "roadway network" location. Knowing that the data collected would eventually be used in a GIS environment, it was paramount that the data collection process be spatially constrained as well as domain constrained. This was not an easy task to perform in a spatially "un-aware" database, so considerable data validation had to be included the front end application. Not doing so could result in erroneous location information needing to be "reworked" at a later time and locations that cannot be referenced against a valid datasource.

Moral of the story.....constrain and validate your data input and never expect the end users to do anything more than "key-in".