No, i think he is referring to a history system.
For each table that you want to track in this way, you will need a 'shadow' or 'history' table next to it to record the changes in.
Say that a table you want to track is
CODE
Table Users:
ID int pk,
Username varchar(250),
RegisterDate DateTime;
Then you historytable could be something like
CODE
Table Users_History:
HistoryID int pk,
ChangeType char(1),
ChangeMoment DateTime,
ID int,
Username varchar(250),
RegisterDate DateTime;
Optionaly you can also add for example a UserID or something to track who was the one that did it.
Note that the primary key from the original table is no longer a primary key. It will even contain duplicates.
To access the reccords in a good fashion a new pk is added that should (also) be a sequence.
Now on the original Users table you will need to make some triggers.
On every insert, update or delete you should insert a reccord into the history table containing the NEW values. (not the old ones, those should already be there from the insert!)
With the delete trigger ofcourse NULLs should be inserted as new values.
I hope this helps.