[tz] Unnecessary faux transition in Atlantic/Canary 1980 Sep/28

Howard Hinnant howard.hinnant at gmail.com
Sat Jun 20 19:45:21 UTC 2015


Hi,

I wouldn’t call this a bug, but a potential database inefficiency.

I’m working on a TZ parser and using it to explore the timezone transitions as specified by the timezone database.  I’m looking at the Atlantic/Canary transition on 1980 Sep 28 0:00s.  At this time the timezone transitions from a permanent saving of 1:00 and abbreviation of WEST to a rule based savings specified by EU:

1980-09-27 23:59:59u = 1980-09-28 00:59:59 WEST // as specified by 0:00 1:00 WEST
1980-09-28 00:00:00u = 1980-09-28 01:00:00 WEST // as specified by 0:00 EU   WE%sT

I.e. there is no change in offset, savings, or abbreviation.  And then an hour later EU specifies a savings and abbreviation change:

1980-09-28 00:59:59u = 1980-09-28 01:59:59 WEST
1980-09-28 01:00:00u = 1980-09-28 01:00:00 WET

My observation is that this is a lot of transitions for one night.  Would it not be better to delay the transition from "0:00 1:00 WEST” to "0:00 EU   WE%sT” by one hour, and thus have only one transition instead of two?  Here is a change that does so:

diff --git a/europe b/europe
index c64c41b..630c234 100644
--- a/europe
+++ b/europe
@@ -2922,7 +2922,7 @@ Zone      Africa/Ceuta    -0:21:16 -      LMT     1901
 Zone   Atlantic/Canary -1:01:36 -      LMT     1922 Mar # Las Palmas de Gran C.
                        -1:00   -       CANT    1946 Sep 30  1:00 # Canaries T
                         0:00   -       WET     1980 Apr  6  0:00s
-                        0:00   1:00    WEST    1980 Sep 28  0:00s
+                        0:00   1:00    WEST    1980 Sep 28  1:00s
                         0:00   EU      WE%sT
 # IATA SSIM (1996-09) says the Canaries switch at 2:00u, not 1:00u.
 # Ignore this for now, as the Canaries are part of the EU.

Howard



More information about the tz mailing list