Thanks Scott. This is very useful insight. I am really only ever concerned with expressing the current timezone for any given place and less worried by any need to relate to any historical dates and times. Consequently it was my hope to use a sanctioned standard that identified timezones as regions with the same political definition (ISO region code), the same UTC offset and the same DST rules but that ignored any historical context; whereby I could find and group all places that exhibited and experienced the same current time behaviour (such as all the cities in China belonging together). But it appears that no such standard exists. Furthermore, and as you point out in the Argentina example, using the "pure" Olsen is recommended in view of the frequent changes that are both "official" as decreed by governments and "unofficial" as adopted by local usage. Having these zones pre-defined as boundary map objects allows us a better degree of future-proofing against changes to come. That is good enough for me. I guess if I need to I might be able to create relationships between zones that behaved the same at any given point in time I might want to look at the CLDR metazones, albeit with some caution. Thanks for your time and help. -Martin ________________________________ From: Scott Atwood [mailto:scott.roy.atwood@gmail.com] Sent: 04 December 2008 17:24 To: tz@elsie.nci.nih.gov Cc: tz@lecserver.nci.nih.gov; Martin Barnes Subject: Re: Time Zone naming To the best of my knowledge, the Olson database itself does not define any kind of "roll-up" timezones. The closest thing I am aware of is the CLDR concept of "metazones" which group together Olson timezones that share a common display string, like "Eastern Standard Time". However, I believe these metazones can include timezones that have different DST rules. Rather than try to use "roll-up" timezones, from personal experience, I would urge you to use the full Olson timezone list if possible. World timezone rules are highly dynamic and change with surprising frequency. And it is not uncommon for two Olson timezones to have the same GMT offset and DST rules in one release of Olson, but then have different rules in a future release. The example that you cite, Argentina, is an excellent example. Until just a few weeks ago, all of Argentina was effective under the same set of time zone rules, but when the central government decided to observe DST this year, several of the states decided to remain in standard time. An application that had assigned "America/Argentina/Buenos_Aires" to everyone in Argentina regardless of their actual Olson timezone would have broken. Also note that it can be useful to maintain the separate timezones if your application needs to format and display historical dates and times, as in logging or transaction history. Timezones that have the same GMT offset and DST rules today may have had different rules in the past, and having the most accurate timezone means you could display the historical records correctly as well. -Scott From: Martin Barnes [mailto:barnes@yahoo-inc.com] Sent: Friday, November 28, 2008 11:48 To: tz@lecserver.nci.nih.gov Subject: Time Zone naming I have a question related to the accepted standard for expressing the "Olsen" name where multiple zones exhibit the same "behaviour" in terms of belonging to the same country, having the same UTC offset and exactly the same DST rules. For example, it appears that all clocks within all locations within Argentina will have the same time all year round. The 12 zones reveal the same behaviour. The same is true of China and a number of other countries. I have been aware of the concept of a "consolidated" or "preferred" time zone which is a combined zone that takes the name of the most important location (eg. "America/Buenos_Aires" in the case of Argentina) Do these combined "super" zones exist? If so, is there information available that indicates how the individual zones roll up? My enquiry relates to a need to provide information that can identify the correct timezone for every place (city, postcode, county, state, etc etc) on earth via a back-end mapping service that calculates the spatial relationship between the place coordinate and the timezone boundary. I am looking to build up an accurate timezone boundary map essentially using existing map objects as building blocks. Many thanks -Martin Barnes ______________________ GeoData Manager Yahoo! Geo Technologies Geo Informatics team London -- Scott Atwood Cycle tracks will abound in Utopia. ~H.G. Wells

Thanks Scott. Yes, I better understand the provenance and status of the Olsen data now. I will look into the details to see the efficacy of constructing the roll-up zones. Since they will be neither an Olsen zone nor a CLDR metazone, we may have issues applying a name. The resultant data will be predominantly Olsen with a few roll-ups but we would want to use the Olsen name throughout. Reflecting on this, I think that it would be prudent to align to the Olsen zones otherwise we might invite problems by associating a place to an Olsen time zone name, when it actually belongs to a zone that has been rolled-up into a "superzone" for which a different name has been applied. I completely understand that the Olsen time zones change from time to time. New ones are added and existing ones may be modified. My question is...is there anyway of being notified when any changes occur to the tz database or is it a matter of systematically checking? -Martin ________________________________ From: Scott Atwood [mailto:scott.roy.atwood@gmail.com] Sent: 04 December 2008 20:05 To: Martin Barnes Cc: tz@elsie.nci.nih.gov; tz@lecserver.nci.nih.gov Subject: Re: Time Zone naming The Olson zoneinfo database and CLDR are the closest thing that presently exist to a standard for timezones, but they are de facto standards, not de jure standards administered by any kind of official international standards body. If it would be useful for your internal purposes, there is certainly enough information in the Olson data to construct "roll-up" timezones for yourself where the ISO-3166-2 code, GMT offset, and DST rules are all presently identical. But as I mentioned below, you should expect such entities to be rather fluid, and you should prefer to base your implementation on the full fidelity Olson timezones. Such "roll-up" timezones would be finer grained entities than CLDR metazones, since the only requirement for metazones is that they must have the same display labels. Individual Olson timezones within a CLDR metazone may have different DST rules and may belong to different regions. Also, on the topic of future proofing, you should be prepared for the fact that new Olson timezones are added from time to time whenever a subregion of an existing zone changes its time definition independently of the rest of that subregion. This happened during the recent timezone change in Argentina where America/Argentina/Salta split off from America/Argentina/Cordoba. Any Olson update could thus potentially require updates to your boundary definitions. -Scott On Thu, Dec 4, 2008 at 10:48 AM, Martin Barnes <barnes@yahoo-inc.com> wrote: Thanks Scott. This is very useful insight. I am really only ever concerned with expressing the current timezone for any given place and less worried by any need to relate to any historical dates and times. Consequently it was my hope to use a sanctioned standard that identified timezones as regions with the same political definition (ISO region code), the same UTC offset and the same DST rules but that ignored any historical context; whereby I could find and group all places that exhibited and experienced the same current time behaviour (such as all the cities in China belonging together). But it appears that no such standard exists. Furthermore, and as you point out in the Argentina example, using the "pure" Olsen is recommended in view of the frequent changes that are both "official" as decreed by governments and "unofficial" as adopted by local usage. Having these zones pre-defined as boundary map objects allows us a better degree of future-proofing against changes to come. That is good enough for me. I guess if I need to I might be able to create relationships between zones that behaved the same at any given point in time I might want to look at the CLDR metazones, albeit with some caution. Thanks for your time and help. -Martin ________________________________ From: Scott Atwood [mailto:scott.roy.atwood@gmail.com] Sent: 04 December 2008 17:24 To: tz@elsie.nci.nih.gov Cc: tz@lecserver.nci.nih.gov; Martin Barnes Subject: Re: Time Zone naming To the best of my knowledge, the Olson database itself does not define any kind of "roll-up" timezones. The closest thing I am aware of is the CLDR concept of "metazones" which group together Olson timezones that share a common display string, like "Eastern Standard Time". However, I believe these metazones can include timezones that have different DST rules. Rather than try to use "roll-up" timezones, from personal experience, I would urge you to use the full Olson timezone list if possible. World timezone rules are highly dynamic and change with surprising frequency. And it is not uncommon for two Olson timezones to have the same GMT offset and DST rules in one release of Olson, but then have different rules in a future release. The example that you cite, Argentina, is an excellent example. Until just a few weeks ago, all of Argentina was effective under the same set of time zone rules, but when the central government decided to observe DST this year, several of the states decided to remain in standard time. An application that had assigned "America/Argentina/Buenos_Aires" to everyone in Argentina regardless of their actual Olson timezone would have broken. Also note that it can be useful to maintain the separate timezones if your application needs to format and display historical dates and times, as in logging or transaction history. Timezones that have the same GMT offset and DST rules today may have had different rules in the past, and having the most accurate timezone means you could display the historical records correctly as well. -Scott From: Martin Barnes [mailto:barnes@yahoo-inc.com] Sent: Friday, November 28, 2008 11:48 To: tz@lecserver.nci.nih.gov Subject: Time Zone naming I have a question related to the accepted standard for expressing the "Olsen" name where multiple zones exhibit the same "behaviour" in terms of belonging to the same country, having the same UTC offset and exactly the same DST rules. For example, it appears that all clocks within all locations within Argentina will have the same time all year round. The 12 zones reveal the same behaviour. The same is true of China and a number of other countries. I have been aware of the concept of a "consolidated" or "preferred" time zone which is a combined zone that takes the name of the most important location (eg. "America/Buenos_Aires" in the case of Argentina) Do these combined "super" zones exist? If so, is there information available that indicates how the individual zones roll up? My enquiry relates to a need to provide information that can identify the correct timezone for every place (city, postcode, county, state, etc etc) on earth via a back-end mapping service that calculates the spatial relationship between the place coordinate and the timezone boundary. I am looking to build up an accurate timezone boundary map essentially using existing map objects as building blocks. Many thanks -Martin Barnes ______________________ GeoData Manager Yahoo! Geo Technologies Geo Informatics team London -- Scott Atwood Cycle tracks will abound in Utopia. ~H.G. Wells -- Scott Atwood Cycle tracks will abound in Utopia. ~H.G. Wells

Thanks Scott Please could I and my colleague, Robert Halliday (emailed here) be added to the tz mailing list? It would be good to have prior notice of upcoming changes and be able to have the opportunity to discuss how time zone boundaries are affected. Many thanks for all your help. -Martin ________________________________ From: Scott Atwood [mailto:scott.roy.atwood@gmail.com] Sent: 05 December 2008 18:11 To: Martin Barnes Cc: tz@elsie.nci.nih.gov; tz@lecserver.nci.nih.gov Subject: Re: Time Zone naming The best way to find about changes to the tz database is to subscribe to and monitor the tz mailing list: tz@elsie.nci.nih.gov. Upcoming changes are typically discussed in advance of their release. -Scott On Fri, Dec 5, 2008 at 9:18 AM, Martin Barnes <barnes@yahoo-inc.com> wrote: Thanks Scott. Yes, I better understand the provenance and status of the Olsen data now. I will look into the details to see the efficacy of constructing the roll-up zones. Since they will be neither an Olsen zone nor a CLDR metazone, we may have issues applying a name. The resultant data will be predominantly Olsen with a few roll-ups but we would want to use the Olsen name throughout. Reflecting on this, I think that it would be prudent to align to the Olsen zones otherwise we might invite problems by associating a place to an Olsen time zone name, when it actually belongs to a zone that has been rolled-up into a "superzone" for which a different name has been applied. I completely understand that the Olsen time zones change from time to time. New ones are added and existing ones may be modified. My question is...is there anyway of being notified when any changes occur to the tz database or is it a matter of systematically checking? -Martin ________________________________ From: Scott Atwood [mailto:scott.roy.atwood@gmail.com] Sent: 04 December 2008 20:05 To: Martin Barnes Cc: tz@elsie.nci.nih.gov; tz@lecserver.nci.nih.gov Subject: Re: Time Zone naming The Olson zoneinfo database and CLDR are the closest thing that presently exist to a standard for timezones, but they are de facto standards, not de jure standards administered by any kind of official international standards body. If it would be useful for your internal purposes, there is certainly enough information in the Olson data to construct "roll-up" timezones for yourself where the ISO-3166-2 code, GMT offset, and DST rules are all presently identical. But as I mentioned below, you should expect such entities to be rather fluid, and you should prefer to base your implementation on the full fidelity Olson timezones. Such "roll-up" timezones would be finer grained entities than CLDR metazones, since the only requirement for metazones is that they must have the same display labels. Individual Olson timezones within a CLDR metazone may have different DST rules and may belong to different regions. Also, on the topic of future proofing, you should be prepared for the fact that new Olson timezones are added from time to time whenever a subregion of an existing zone changes its time definition independently of the rest of that subregion. This happened during the recent timezone change in Argentina where America/Argentina/Salta split off from America/Argentina/Cordoba. Any Olson update could thus potentially require updates to your boundary definitions. -Scott On Thu, Dec 4, 2008 at 10:48 AM, Martin Barnes <barnes@yahoo-inc.com> wrote: Thanks Scott. This is very useful insight. I am really only ever concerned with expressing the current timezone for any given place and less worried by any need to relate to any historical dates and times. Consequently it was my hope to use a sanctioned standard that identified timezones as regions with the same political definition (ISO region code), the same UTC offset and the same DST rules but that ignored any historical context; whereby I could find and group all places that exhibited and experienced the same current time behaviour (such as all the cities in China belonging together). But it appears that no such standard exists. Furthermore, and as you point out in the Argentina example, using the "pure" Olsen is recommended in view of the frequent changes that are both "official" as decreed by governments and "unofficial" as adopted by local usage. Having these zones pre-defined as boundary map objects allows us a better degree of future-proofing against changes to come. That is good enough for me. I guess if I need to I might be able to create relationships between zones that behaved the same at any given point in time I might want to look at the CLDR metazones, albeit with some caution. Thanks for your time and help. -Martin ________________________________ From: Scott Atwood [mailto:scott.roy.atwood@gmail.com] Sent: 04 December 2008 17:24 To: tz@elsie.nci.nih.gov Cc: tz@lecserver.nci.nih.gov; Martin Barnes Subject: Re: Time Zone naming To the best of my knowledge, the Olson database itself does not define any kind of "roll-up" timezones. The closest thing I am aware of is the CLDR concept of "metazones" which group together Olson timezones that share a common display string, like "Eastern Standard Time". However, I believe these metazones can include timezones that have different DST rules. Rather than try to use "roll-up" timezones, from personal experience, I would urge you to use the full Olson timezone list if possible. World timezone rules are highly dynamic and change with surprising frequency. And it is not uncommon for two Olson timezones to have the same GMT offset and DST rules in one release of Olson, but then have different rules in a future release. The example that you cite, Argentina, is an excellent example. Until just a few weeks ago, all of Argentina was effective under the same set of time zone rules, but when the central government decided to observe DST this year, several of the states decided to remain in standard time. An application that had assigned "America/Argentina/Buenos_Aires" to everyone in Argentina regardless of their actual Olson timezone would have broken. Also note that it can be useful to maintain the separate timezones if your application needs to format and display historical dates and times, as in logging or transaction history. Timezones that have the same GMT offset and DST rules today may have had different rules in the past, and having the most accurate timezone means you could display the historical records correctly as well. -Scott From: Martin Barnes [mailto:barnes@yahoo-inc.com] Sent: Friday, November 28, 2008 11:48 To: tz@lecserver.nci.nih.gov Subject: Time Zone naming I have a question related to the accepted standard for expressing the "Olsen" name where multiple zones exhibit the same "behaviour" in terms of belonging to the same country, having the same UTC offset and exactly the same DST rules. For example, it appears that all clocks within all locations within Argentina will have the same time all year round. The 12 zones reveal the same behaviour. The same is true of China and a number of other countries. I have been aware of the concept of a "consolidated" or "preferred" time zone which is a combined zone that takes the name of the most important location (eg. "America/Buenos_Aires" in the case of Argentina) Do these combined "super" zones exist? If so, is there information available that indicates how the individual zones roll up? My enquiry relates to a need to provide information that can identify the correct timezone for every place (city, postcode, county, state, etc etc) on earth via a back-end mapping service that calculates the spatial relationship between the place coordinate and the timezone boundary. I am looking to build up an accurate timezone boundary map essentially using existing map objects as building blocks. Many thanks -Martin Barnes ______________________ GeoData Manager Yahoo! Geo Technologies Geo Informatics team London -- Scott Atwood Cycle tracks will abound in Utopia. ~H.G. Wells -- Scott Atwood Cycle tracks will abound in Utopia. ~H.G. Wells -- Scott Atwood Cycle tracks will abound in Utopia. ~H.G. Wells