I should explain what I meant by "misleading". Back to 1970, you know that you can find a tzid matching any place, and that tzid will produce the right results for that place at all times back to 1970. Before 1970, the situation is a bit different. You are not guaranteed that there will be a tzid that captures the precise behavior for every place. You know what each tzid means, but you don't when a tzid should have been broken into multiple ones to reflect differences in behavior that existed back then. By analogy, suppose that the cutoff point were 1990 instead of 1970. There might be only one tzid for, say, China. If you used that tzid with a time that happened before 1990, for someplace like Harbin, you'd get the wrong results and not realize it. Now the significance of this "completeness" depends on how important it is to get the conversions correct. If I have a calendar program that communicates a weekly recurring meeting, and communicates that to other participants with the internal information <08:00, America/Los Angeles>, then I can depend on the participants' programs being able to correctly show on their calendars precisely when that is in their local time, which may change from week to week. It is clearly extremely important to get current times right, and near-future and near-past times. For specialized applications, it is probably also important over longer periods; a large database of data like FBI records might be an example. But for most general programs it is not as important to get it all right back before 35 years ago, so I think 1970 is a reasonable cut-off for the developers of the database to guarantee that the tzids are complete. Mark ----- Original Message ----- From: "Mark Davis" <mark.davis@jtcsv.com> To: "David Keegel" <djk@cybersource.com.au> Cc: "Funda Wang" <fundawang@gmail.com>; "Paul Eggert" <eggert@CS.UCLA.EDU>; "Tz (tz@elsie.nci.nih.gov)" <tz@lecserver.nci.nih.gov> Sent: Saturday, February 05, 2005 16:02 Subject: Re: Re[2]: Corrections to timezone database
True, and I should have mentioned that. The database is a bit misleading in that it does provide for data before that point for tzids, even though logically some of them should be broken into several IDs based on differences before 1970. On the other hand, one can well understand putting a cut-off point in place, if only because it is difficult to get reliable data earlier, and it is not as important for implementations.
Mark
----- Original Message ----- From: "David Keegel" <djk@cybersource.com.au> To: "Mark Davis" <mark.davis@jtcsv.com> Cc: "Funda Wang" <fundawang@gmail.com>; "Paul Eggert" <eggert@CS.UCLA.EDU>; "Tz (tz@elsie.nci.nih.gov)" <tz@lecserver.nci.nih.gov> Sent: Saturday, February 05, 2005 15:54 Subject: Re: Re[2]: Corrections to timezone database
On Sat, Feb 05, 2005 at 02:55:32PM -0800, Mark Davis wrote:
As far as we are concerned, the tzid is simply an internal tag,
marking a
region of the Earth that has the same timezone behavior, all the way back to the point in time where timezones started to be used instead of solar time.
The Theory file says: * Uniquely identify every national region where clocks have all agreed since 1970.
So change "all the way back to the point in time where timezones started to be used instead of solar time" to "since the year 1970."
___________________________________________________________________________
David Keegel <djk@cybersource.com.au> http://www.cyber.com.au/users/djk/ Cybersource P/L: Linux/Unix Systems Administration Consulting/Contracting