[tz] Greetings from Mongolia

[Paul Eggert]

amgalan tsogtgerel wrote:

> the CRC has sent you official letter dated  15th June 2015 to you for your support on above mentioned issue and we  have  not received any response from you yet.

I twice replied via email to crc at mongol.net, the email address mentioned in that 
official letter.  Perhaps you're not getting those emails?  Anyway, at the risk 
of repeating myself:

Although the tz project and the IANA are doing their job, Apple and Google and 
etc. have messed up in Mongolia.  I suggest that you write about this to Apple 
and Google and etc.; possibly you'll have more influence on them than I do. 
Please feel free to quote this email, particularly paragraph (B) below.  I 
suggest also writing to the part of the Mongolian government responsible for 
daylight-saving time and mentioning paragraph (A).

Here are more details about the issue.

First, the Mongolian time change works for me on my desktop, which is running 
Ubuntu 15.04.  I can run the following programs from my shell:

$ TZ=Asia/Ulaanbaatar date; date -u
Thu Jul  2 00:47:32 ULAST 2015
Wed Jul  1 15:47:32 UTC 2015

This correctly says Mongolian time is Ulaanbaatar Summer Time (ULAST), which is 
nine hours ahead of Coordinated Universal Time (UTC).  It's also working that 
way for me on my main UCLA cluster, running CentOS 6.0 (a Red Hat clone released 
in 2011).  Also, on another UCLA-based server, running Oracle Solaris 11.2 
(released 2014).

So it is working in some cases.  But not all.  I found one old server at work, 
running Solaris 10 (released 2005), which says Mongolian time is only eight 
hours ahead of UTC; this server's time zone tables haven't been updated since 
May 2012.  More importantly, my son's cell phone and my cell phone have similar 
problems; they're running iOS 8.3 and Android 5.1.1 (both are current versions 
released April 2015); apparently Apple and Google did not update their operating 
systems' time zone tables before releasing them and have not patched them since 
release.  Undoubtedly this is what Mongolian users are complaining about.


Before talking about how to fix things, let me describe how the process worked 
and/or is not working:

1.  The Mongolian government decided on 2015-03-09 to use daylight saving time 
starting 2015-03-28 03:00 local time.

2.  Ganbold Tsagaankhuu (affiliation unknown) privately sent me email dated 
2015-03-10 notifying me of the change.

3.  We published an experimental patch dated 2015-03-10 00:03:48 -0700 
containing the change; you can see its archived announcement at 
<http://mm.icann.org/pipermail/tz/2015-March/022073.html>.

4.  We published an official release 2015b dated 2015-03-19 23:28:11 2015 -0700 
containing the change; you can see its archived announcement at 
<http://mm.icann.org/pipermail/tz-announce/2015-March/000029.html>.

5.  Suppliers of operating systems eventually pick up these changes and release 
new operating system releases and/or patches to their existing systems.  Known 
operating systems include GNU/Linux distributions like Red Hat and Ubuntu, and 
many other systems including Android (Google), Firefox OS, FreeBSD, NetBSD, 
OpenBSD, Cygwin, DJGPP, MINIX, webOS (LG), AIX (IBM), BlackBerry 10, iOS 
(Apple), Microsoft Windows, OpenVMS (HP), Oracle Database, Oracle Solaris, and 
OS X (Apple).  (There are undoubtedly other operating systems that we don't know 
about.)  From the above examples, we know that Android and iOS have not been 
updated, whereas Red Hat and Solaris and Ubuntu have been.

6.  Users of these operating systems eventually install patches to bring their 
systems up to date.  For proprietary systems this often requires support contracts.

Steps (5) and (6) haven't been completed properly in Mongolia; that is, the 
Mongolian systems in question are like my older server at work (where step (6) 
hasn't been done), or are like Apple or Android cell phones (where step (5) 
hasn't been done).


Here are some ways to improve this process the next time the Mongolian 
government changes its time zone rules.

A.  The Mongolian government can announce the change well before it occurs. 
Eighteen days is not enough.  I suggest six months or more.  The last time the 
rules were changed in the US, the federal government gave nineteen months' notice.

B.  Operating system suppliers can be more timely about propagating the changes 
to their operating systems.  My desktop contains all the changes of tz release 
2015d (dated April 24); why doesn't my cell phone?  Apple and Google should have 
the current tz version in their current cell phone releases and should send 
over-the-air patches in a timely way.

C.  Operating system users in Mongolia can be more diligent about installing 
patches.  Users regularly employing over-the-network updates (which is good 
practice anyway) should have no problems.

D.  The Internet Engineering Task Force has drafted the specifications for a 
time zone data distribution service that should automate steps (B) and (C) more 
rapidly.  See
<https://tools.ietf.org/html/draft-ietf-tzdist-service-09> (2015-06-29).  My 
impression is that Apple will implement this service, and perhaps other 
suppliers will do something similar.  However, it's just a draft and is not 
widely implemented so this is somewhat speculative.


PS.  This problem is not unique to Mongolia.  We have similar problems in 
Uruguay later this year (fix not yet officially released), in Egypt this year 
(fixed in tz release 2015d dated 2015-04-24 08:09:46 -0700), in Chile and Mexico 
this year (fixed in tz release 2015a dated 2015-01-29 22:35:20 -0800), etc.