On 2018-08-17 13:13, Tom Lane wrote:
Howard Hinnant <howard.hinnant@gmail.com> writes:
On Aug 17, 2018, at 3:53 AM, Paul Eggert <eggert@cs.ucla.edu> wrote:
"The biggest hurdle in Japan to putting daylight saving time in place is the cost and workload required to adjust computer systems. Professor Tetsutaro Uehara of Ritsumeikan University, a specialist in information systems, estimates that it would take about four years and hundreds of billions of yen to do just that. Let’s send ‘em a bill. ;-) I rather imagine the professor has a point. Yeah, code that uses tzdb would be easy to update. But Japan is likely chock-full of locally grown code that has never had to cope with any situation other than "JST = UTC+9", and probably hasn't got any generality whatsoever about its timestamp handling.
I doubt the bewildered professor has enough background in industrial, commercial, and retail software development and deployment practices, infrastructure and apps to make a reasonable estimate. Recall the posts we've had about squishing and updating tzdbs in firmware: major embedded libraries support tzdb and locales (although they can usually be omitted), get contributions from the chip makers who use them, and BSD and TZ code can be freely leveraged to support DST. As usual, the main problem will be updating or replacing software and systems not designed to be updated because of assumptions: like pre-Y2K, and 2005-2007 for the North American DST dates rule changes. -- Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada