Here is a way to use Oracles internal knowledge of whether a timezone observes daylight saving time or not to determine the start and end of it. Aside from the complexity and general strangeness of it, it requires two timezones to be know have identical times when daylight saving time is not in effect and different times when it is. As such it is resilient to congressional changes in when daylight saving time occurs (assuming your database is up to date with the patches), but is not resilient to regional changes effecting the timezones keyed off of. With those warnings, here is what I have.
ALTER SESSION SET time_zone='America/Phoenix';
DROP TABLE TimeDifferences;
CREATE TABLE TimeDifferences(LocalTimeZone TIMESTAMP(0) WITH LOCAL TIME ZONE);
INSERT INTO TimeDifferences
(
SELECT to_date('01/01/' || to_char(sysdate-365,'YYYY') || '12:00:00','MM/DD/YYYYHH24:MI:SS')+rownum-1
FROM dual CONNECT BY rownum<=365
);
COMMIT;
ALTER SESSION SET time_zone='America/Edmonton';
SELECT LocalTimeZone-1 DaylightSavingTimeStartAndEnd
FROM
(
SELECT LocalTimeZone,
to_char(LocalTimeZone,'HH24') Hour1,
LEAD(to_char(LocalTimeZone,'HH24')) OVER (ORDER BY LocalTimeZone) Hour2
FROM TimeDifferences
)
WHERE Hour1 <> Hour2;
I told you it was strange. The code only figures out the day of the change, but could be enhanced to show the hour. Currently it returns 09-MAR-08 and 02-NOV-08. It is also sensitive to the time of year it is run, which is why I had to do the -365...+365. All in all I don't recommend this solution, but it was fun to investigate. Maybe someone else has something better.