I heartily agree with Beth here - but with justifications as to why it should be done as opposed to as a last resort:
Data should be stored in its native format. You should no more store it as a string than you should an 8-bit integer. It makes just as much/little sense. SQL is a very intelligent beast, but if you screw up your fundamental data types most of its useful functionality and heuristics goes out of the window - even basic stuff like indexing and sorting. Not to mention your data becomes SQL Server explicit (ie. relies on certain SQL behaviour) as opposed to generic, reducing portability.
To put it in more formal terms: you're forcing it to become information at the data layer. This is antithesis to all known rules of data handling.