This is not a joke. A company I know of has actually decided to build a large production database where all of the data will be place in one large varchar column, of one huge table. There is another table that contains a "Label" and row type id that can be used to find the item in the other table. They say this will allow them to add new data items to the system without having to update the schema.
Among the issues I see are:
There is no way to index anything.
Querying the data is a nightmare and requires joining 7-8 tables together to get one item that would be a column in a more traditional database.
Because all data is stored in character form, doing any compares will first require that the entire data set you are looking at be converted to the correct type. (What happens when data that is supposed to be a date does not convert without an error?)
Performance will be terrible.
I have used this approach before, for certain limited cases. For example when storing the questions and answers to surveys, where the questions change from one year to the next.
I am just wondering what industry professionals thing about this.