daylight-saving time in Egypt; problem with zic and `24:00'

[Paul Eggert]

   From:	Rives McDow [SMTP:rmcdow at ibm.net]
   Sent:	Thursday, September 24, 1998 9:56 AM

   Egypt changes their time at midnight between Thursday and Friday,
   so when you say the time changes the 4th Friday in September, are
   you meaning this?

My intent was to say that the time changed at 24:00 on the last Friday
in September.  However, I see that I entered that data incorrectly.
Thanks for pointing this out.

Here's the problem: unfortunately, `zic' does not let you enter the
transition time `24:00', meaning midnight at the end of the day; it
complains `invalid time of day'.  `zic' should probably be enhanced to
allow 24:00, as that is often what the rules actually say -- Egypt is
an example of this.

I work around this problem by specifying 00:00 on the next day.  This
is tricky for Egypt, because `Sep lastFri 00:00' is not equivalent to
`Sep lastThu 24:00'; e.g. the former specifies the incorrect value
1999-09-24 00:00 whereas the latter specifies the correct value
1999-10-01 00:00.  So I intended to cheat by writing `Sep lastThu
23:00s', which happens to work because the transition occurs during
DST when 23:00s is equivalent to 24:00.  Unfortunately, I slipped and
wrote `lastFri' instead of `lastThu'.

The following test patch to `africa' should address this problem;
could you please try it out?

--- africa	1998/09/24 14:46:08	1998.8
+++ africa	1998/09/25 20:29:50
@@ -208,7 +208,7 @@ Rule	Egypt	1989	only	-	May	 6	1:00	1:00	
 Rule	Egypt	1990	1994	-	May	 1	1:00	1:00	S
 # IATA (after 1990) says transitions are at 0:00.
-# Go with IATA starting in 1995.
+# Go with IATA starting in 1995, except correct 1995 entry from 09-30 to 09-29.
 Rule	Egypt	1995	max	-	Apr	Fri>=22	 0:00s	1:00	S
-Rule	Egypt	1995	max	-	Sep	lastFri	23:00s	0	-
+Rule	Egypt	1995	max	-	Sep	lastThu	23:00s	0	-
 
 # Zone	NAME		GMTOFF	RULES	FORMAT	[UNTIL]


   As far as the 4th Friday as opposed to the last Friday, my sources
   have told me that it is the last Friday, but I am having them check
   more thoroughly to clear this up.

Thanks for looking into this; please let me know how your researches
come out, and I'll include the results in my next proposed patch.  If
the spring transition is indeed the 24:00 on last Thursday in April,
then I guess I'll use the even kludgier cirumlocution `Apr lastThu 22:00u'.

Here is the IATA data for Egypt's DST transitions since 1991, as
provided to me by Gwillim Law on Aug 6.  All times are UTC.  I think
the `Sep 29 1995' entry is a typo, and should be `Sep 28 1995'; if you
could double-check this too, I'd appreciate it.  The 1998 and 1999
entries both suggest something other than a lastThu rule for April,
though it's hard to be sure exactly what the rule should be, and my
`Apr Fri>=22 0:00s' is pretty much guesswork.

Apr 30 1991 10:00PM	Sep 30 1991 9:00PM
Apr 30 1992 10:00PM	Sep 30 1992 9:00PM
Apr 30 1993 10:00PM	Sep 30 1993 9:00PM
Apr 30 1994 10:00PM	Sep 30 1994 9:00PM
Apr 27 1995 10:00PM	Sep 29 1995 9:00PM
Apr 25 1996 10:00PM	Sep 26 1996 9:00PM
Apr 24 1997 10:00PM	Sep 25 1997 9:00PM
Apr 23 1998 10:00PM	Sep 24 1998 9:00PM
Apr 22 1999 10:00PM	Sep 30 1999 9:00PM