Which DML statements create undo information?

Question

I have a quickie here - the reason I need to clarify is that there is a contradiction between the Couchman Fundamentals text book, and the CD.
CD says : DML that creates undo are DELETE, INSERT, UPDATE
Book explains: it is only DELETE, UPDATE.
Which to believe, or is there a third way ?
DA

Answer 1

Yeahh as I've added my voice to mentioned thread by yours, I've repeat the example here also abot undo:
(rows with 0 values I deleted)
SQL> select version from v$instance;
VERSION
10.2.0.1.0
SQL> select value, name
  2  from v$statname n, v$mystat m
  3  where n.statistic# = m.statistic#
  4    and (upper(name) like '%REDO%'
  5     or upper(name) like '%UNDO%')
  6  /
     VALUE NAME
         3 redo synch writes
        72 redo entries
     18260 redo size
      5536 undo change vector size
         5 data blocks consistent reads - undo records applied
       372 IMU undo allocation size
33 rows selected.
SQL> with a as (select /*+ materialize */ * from dual)
  2  select * from a;
D
X
SQL> select value, name
  2  from v$statname n, v$mystat m
  3  where n.statistic# = m.statistic#
  4    and (upper(name) like '%REDO%'
  5     or upper(name) like '%UNDO%')
  6  /
     VALUE NAME
         4 redo synch writes
        79 redo entries
     19808 redo size
      5780 undo change vector size
         5 data blocks consistent reads - undo records applied
       372 IMU undo allocation size
33 rows selected.Unfortunately in 9i there isn't such metric like "undo change vector size".
Gints Plivna
http://www.gplivna.eu