sys.index_columns.key_ordinal versus index_column_id

http://msdn.microsoft.com/en-us/library/ms175105.aspx

What is the difference between key_ordinal and index_column_id?

One one table where there is no PK indicated by there is a clustered unique index with four fields, the key_ordinal values reflect the order of the fields in the index, the index_column_id has 3-4 reversed.

So I mainly wonder what the purpose of index_column_id is?

Thanks.

Josh

ps - SQL 2008 64-bit enterprise.