Why do we use Meaningful PK or Meaningless PK ?

Question

Hi All,
My questions below is very basic (I am a developer not a dba ..)
I see on database design that the Primary Key (PK) is usually not the "Key" that is visible to user, e.g : if we has Invoice table, the columns are :
InvoiceID (PK) => this is only visible to developer, user never see it, usually
using sequence.
InvoiceNumber => this is visible on UI, reports etc
From the design point of view, what is the reason of the above design ?
Thank you for your help,
xtanto

Answer 1

Guys, my opinions are as follows...
1) Psychology says that human being like the things in order. If that thing is primary key (with a sequence) of your table then, it will help you alot in one or the other way.
2) I feel very comfortable and confident when I write complex queries using surrogate keys.
3) Suppose I say I will use natural keys for parent tables. In some situations, child tables has to be implemented by surrogate keys. This will result in some tables with natural keys and others with surrogate keys. This may sometimes lead to confusion. To avoid this type of situation and to maintain consistency, people prefer to use surrogate keys in all the tables.
4) Surrogate keys help in guessing. Let us see the following rows (first column is a key)....
With surrogated keys
Key, bank_account_number, customer_name,description
456,123765,'oracle corporation','business sector'
457,409867,'rotary high school','education sector'
458,6784,'AT and T','telecommunication sector'
With Natural keys
bank_account_number, customer_name,description
123765,'oracle corporation','business sector'
409867,'rotary high school','education sector'
6784,'AT and T','telecommunication sector'
In both the cases, the first column is a primary key. With surrogate keys, I can easily say 'AT and T' got entered in the table after 'rotary high school' and 'oracle corporation'. You remember the key 458 for 'AT and T' and assume you have millions of such rows. In your development sometimes you will come across with a row with 123. Surrogate key helps you to say 'Oh this is old row than AT and T row'. May be your GM asked. "Dear DBA which customer is old, by providing a list of customers?" . I just have to follow the surrogate keys. In the second case you will not have that type of implicit guessing.
5) Assume the cases as in point 4. May be 409867 account number of 'rotary high school' of bank 'US federal bank' matches with 409867 account number of other customer say 'xyz' of bank 'Australia federal bank'. In this case, you will think to put (account number,bank name) as primary key. In some other cases you will have to have 3 or 4 columns to uniquely identify the row. I think instead of including primary key with two or more columns, why not use surrogate key with just one column. This surrogate key never cares about the data present in other collegue columns in the table.
Hope I put some light on this issue. I welcome others to help me understand the importance of surrogate keys in a more better way.
-aijaz