<div dir="ltr"><div dir="ltr"><div>Formally speaking, Ethiopia is already "supported" by tz insofar as its offset of UTC+3 is recorded and referenced as Africa/Addis_Ababa (which is a link to Africa/Nairobi). But, even if we ignore the differences between atomic and solar time — it sounds like both see some use, depending on the precision required — the unique way of naming the hours is a different problem entirely.</div><div><br></div>Much like we're seeing with current discussion re: fallback transitions in mid-20th-century Japan, the tz database isn't really concerned so much with what the hours are <i>called</i> as how they are <i>counted</i>… Or at least, that's the best I can say in words; it's a wiggly distinction, to be sure, and edge cases like this do exist. All tz does, though, given a timestamp, is to convert that into some local reckoning by spitting out the elements of a Gregorian date and a time-of-day according to a standard 24-hour clock, counting from 00:00:00. Obviously, that intuition gets hairy at various time discontinuities (DST and leap seconds are both examples, but are not handled analogously). However, convention has been established by some decades of global precedent and practice. While it would be technically possible to implement different models for this, it would likely require some major code changes.</div><div dir="ltr"><br></div><div dir="ltr">In most of the world, tz's output maps pretty cleanly to how time is reckoned by everyday folk: e.g., right now, we're I'm at, it's 17:00. In some parts, there may be a conventional mapping to other customary terminology; e.g., it's 5:00pm. This 12-hour system maps what tz calls "hour 0" to the concept of "12am", and so on through "hour 23" being called "11pm". And while it is widely supported in most systems because it is a relatively simple transformation and its use is prevalent throughout the world (though a historical dose of US-centrism likely helped, too), it has nothing to do with tz's outputs themselves, and more to do with how those results are formatted. In many parts of Europe and Asia, 17:00 could be used directly and would be well-understood.</div><div dir="ltr"><br></div><div dir="ltr">This is analogous to mapping dates to different calendar systems: tz uses the proleptic Gregorian calendar and will tell me, in its structured data format, that today is 2018-09-12, and while that's directly useful to telling me it's "12 September 2018" in the calendar I use, if I wanted to know that it was "3 Tishrei 5779" in the Hebrew calendar or "1 Muharram 1440" in the Islamic calendar, I would need to perform a calculation, and that calculation is outside the scope of tz.<br><div><br></div><div>The gut reaction of some might be to paper over Ethiopia's unique system by picking some timezone 6 hours east or west (UTC–3 or UTC+9, for example), but it raises additional questions: When should the date change over, near dusk, near dawn — or still at "midnight", which is to say "6 o'clock"? Does traditional Ethiopian timekeeping have the same notion of "am" or "pm", and is that even accurate terminology, if we were to translate that notion into English? Does Ethiopian practice match the Western convention which starts the "am" and "pm" periods with the hour called "12" and not with "1"?</div><div><br></div><div>I'd be inclined to say that the "correct" way to support the traditional Ethiopian timescale would be for tz to still output Ethiopian time in the standardized way it already does, and for applications and operating systems to then understand that "17:00" in the standard system doesn't mean "5 o'clock", but rather should be displayed to be understood as "11 o'clock". It could be another localization option, just like alternate calendars are today. Unfortunately, unlike the various calendar systems, which enjoy a decent amount of support, Ethiopia is (to our knowledge) unique in this regard. But a decent thought experiment would be to consider how tz and localization projects like CLDR would have managed with <a href="https://en.wikipedia.org/wiki/Decimal_time">Decimal Time</a> in the French Revolution. That's a <i>very</i> different way of counting hours; at least Ethiopia's is relatively similar to things we already support.</div><div><br></div><div>Depending on how seriously others take, say, the prospect of supporting other timescales entirely (TAI or GPS, perhaps, but maybe as far-reaching as Mars time) becoming important in the future, the place for that sort of support might land in that project, be it tz or otherwise. But for now, I think it's more of a localization issue than a timekeeping one: They're using different names for the same underlying thing.</div><div><br></div></div><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature">--<br>Tim Parenti<br></div></div><br><br><div class="gmail_quote"><div dir="ltr">On Wed, 12 Sep 2018 at 16:06, <<a href="mailto:Paul.Koning@dell.com">Paul.Koning@dell.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
<br>
> On Sep 12, 2018, at 3:58 PM, Steve Allen <<a href="mailto:sla@ucolick.org" target="_blank">sla@ucolick.org</a>> wrote:<br>
> <br>
> On Tue 2018-09-11T22:23:01-0700 Paul Eggert hath writ:<br>
>> I'm afraid not, as it's based on solar time.<br>
> <br>
> The article gives the impression that it's not really solar because an<br>
> hour of imprecision is culturally irrelevant, so it's really just 6<br>
> hours different from what a cell phone says. But even with that<br>
> simplicity I don't think that tz should implement something without a<br>
> much more authoritative and clear source for when the Ethiopian day<br>
> begins.<br>
> <br>
> --<br>
> Steve Allen <<a href="mailto:sla@ucolick.org" target="_blank">sla@ucolick.org</a>> WGS-84 (GPS)<br>
> UCO/Lick Observatory--ISB 260 Natural Sciences II, Room 165 Lat +36.99855<br>
> 1156 High Street Voice: +1 831 459 3046 Lng -122.06015<br>
> Santa Cruz, CA 95064 <a href="http://www.ucolick.org/~sla/" rel="noreferrer" target="_blank">http://www.ucolick.org/~sla/</a> Hgt +250 m<br>
<br>
That brings up an interesting question.<br>
<br>
What does the TZ database mean "local time" to be? Time in its common representation where zero is midnight and 12 o'clock is noon? Or is it meant to account also for local conventions that zero is some point in the day different from midnight?<br>
<br>
If the former, then this issue is out of scope. If the latter, then it suggests there might be two Ethiopia zones, one for "midnight origin" (the one we have now) and one for "local convention" which combines the offsets from latitude, and the offset from the different convention of what the starting point is.<br>
<br>
paul<br>
</blockquote></div></div>