I'm curious to whether you'd recommend this approach for a particularly gnarly database I've inherited. It uses a concept of "partial text" for it's keys. For example, 2007-000123-001 might be the key in a table. The first 11 characters is, in essence, the primary key of the parent table in a parent-child hierarchy. The -001 makes it unique in the child table and also gives it a sequential order.
As you can imagine, it's painful because searches invariably rely upon substrings or LIKE clauses. I know that's killing the performance.
My thought was that I could use this computed column approach to slowly migrate us out of this hell, while still maintaining legacy support. I'd be curious to hear your thoughts.