If I use a command that contains field name aliases, I get a different
result for _cmdBldr.GetUpdateCommand().CommandText, depending on
whether I am connected to SQL Server 2005 or Oracle 10g.

Here is the code in question:

    string _sqlString = "SELECT ID, PLANGROUP AS PG, DESC_R FROM TABLEA
WHERE ID = 100002";
    OdbcConnection _dc = new OdbcConnection();
    _dc.ConnectionString = "<connectionstring>";
    _dc.Open();
    OdbcDataAdapter _da = new OdbcDataAdapter(_sqlString, _dc);
    DataTable _dt = new DataTable();
    _da.Fill(_dt);
    _dt.Rows[0].BeginEdit();
    _dt.Rows[0]["DESC_R"] = "ABC";
    _dt.Rows[0].EndEdit();
    OdbcCommandBuilder _cmdBldr = new OdbcCommandBuilder(_da);
    MessageBox.Show(_cmdBldr.GetUpdateCommand().CommandText);
    try
    {
        _da.Update(_dt);
    }
    catch(Exception ex)
    {
    }

When I run this for a SQL Server connection, the _da.Update(_dt);
command executes without an exception and
_cmdBldr.GetUpdateCommand().CommandText returns:

"UPDATE TABLEA SET ID = ?, PLANGROUP = ?, DESC_R = ? WHERE ((ID = ?)
AND ((? = 1 AND PLANGROUP IS NULL) OR (PLANGROUP = ?)) AND ((? = 1 AND
DESC_R IS NULL) OR (DESC_R = ?)))"

When I run this for an Oracle connection the _da.Update(_dt); command
generates an exception:

"ERROR [42S22] [Oracle][ODBC][Ora]ORA-00904: "PG": invalid identifier"

and _cmdBldr.GetUpdateCommand().CommandText returns:

"UPDATE TABLEA SET ID = ?, PG = ?, DESC_R = ? WHERE ((ID = ?) AND ((?
= 1 AND PG IS NULL) OR (PG = ?)) AND ((? = 1 AND DESC_R IS NULL) OR
(DESC_R = ?)))"

Note that in SQL Server, the original field name "PLANGROUP" is used
in the UpdateCommand.  In Oracle, the field name alias "PG" is used.

Is there a way to force the command builder to create an update
command with the orginal field name when using Oracle?  Is there
another approach I should be using, other than always creating my own
update command from scratch for Oracle?

Thanks.