On Sep 5, 2013, at 12:06 PM, Lester Caine <lester@lsces.co.uk> wrote:

It is important to point out that neither of this are accurate around the edges. In many places due to the simplified shapes the wrong timezone is returned so one has to take care.

Have you submitted corrected shapefiles, or suggested corrections to the shapefiles, to Eric Muller? http://efele.net/contact.html

On Sep 5, 2013, at 12:30 PM, Lester Caine <lester@lsces.co.uk> wrote:

Zoom in on http://maps.derickrethans.nl/?l=timezone,timezones&lat=52.485&lon=-1.890&zoo... the entire database of shape files needs to be replaced with the administrative boundaries from the OSM database,

To which *level* of administrative boundary are you referring? http://wiki.openstreetmap.org/wiki/Tag:boundary%3Dadministrative Level 2 (nation-state boundaries) is obviously insufficient to handle the entire tzdb, and going down to, for example, the US's level 8 ("municipalities, including cities, villages, etc. Also incorporated townships (New Jersey, Pennsylvania.)") is equally obviously overkill in most cases. In order to use OSM to construct boundaries for tzdb zones, a *mixture* of administrative boundaries would be required, so that *wouldn't* constitute replacing tzdb boundaries with administrative boundaries from the OSM database, it would constitute *constructing* higher-quality tzdb boundaries from the administrative boundaries.

Guy Harris wrote:

And if there are any tzdb zones that correspond to*historical* boundaries not represented in the current OSM, you're going to get boundaries from the 4D version of the OSM: http://wiki.openstreetmap.org/wiki/Historical

http://www.openhistoricalmap.org/ is rather empty at the moment, but yes changes to boundary locations are already being archived in the change log, and adding historic areas will be catered for in the OHM. -- Lester Caine - G8HFL ----------------------------- Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

On Sep 5, 2013, at 1:54 PM, Lester Caine <lester@lsces.co.uk> wrote:

The reason that an adminastrative boundary for timezone is not yet available is because nobody has had time to create it. It will use elements of existing relations, but to be honest the current thinking is that simply tagging elements that have a unique timezone with that information. That way when one looks up a place record you get the information without having to do a geometric search.

If all you have is longitude and latitude, you're going to have to do a geometric search. There might not *be* a place record for where you are, and, even if there is, it won't necessarily have a tag. They claim to have 105 places tagged with America/New_York; to quote https://en.wikipedia.org/wiki/Administrative_divisions_of_New_York "New York has 62 counties, which are subdivided into 932 towns and 62 cities;[4] it also has 10 Indian reservations.", so that's about 10.5% of New York State, and America/New_York also includes: Connecticut Delaware Georgia Maine Maryland Massachusetts New Hampshire New Jersey North Carolina Ohio Pennsylvania Rhode Island South Carolina Vermont Virginia West Virginia and parts of some other states as well. They probably have the major cities covered, but 1) there are probably plenty of UN*X boxes using the tzdb outside of those major cities and 2) the vast majority of smartphones sold in the US are probably UN*X boxes using the tzdb (I'm guessing Bionic uses the tzdb; Darwin's libSystem definitely does), and they can be on the road *between* places.

On Sep 5, 2013, at 2:27 PM, random832@fastmail.us wrote:

On Thu, Sep 5, 2013, at 17:16, Guy Harris wrote:

...

If all you have is longitude and latitude, you're going to have to do a geometric search. There might not *be* a place record for where you are, and, even if there is, it won't necessarily have a tag. They claim to have 105 places tagged with America/New_York; to quote

https://en.wikipedia.org/wiki/Administrative_divisions_of_New_York

I don't know why you're quoting that article.

I'm quoting that article because the city with which the tzid is associated is within that state, and fewer neurons had to fire than to choose, for example, the state in which I was born or the state in which I went to institute^Wuniversity. I'm not quoting any *other* articles because I'm too damn lazy to quote the articles for all the states I listed.

I suspect the majority of locations tagged with America/New_York are A) states that are entirely (or almost entirely) in eastern time and B) counties in states that are on the border between it and central time.

Most of them are "relations": http://wiki.openstreetmap.org/wiki/Relation and are probably either: http://wiki.openstreetmap.org/wiki/Types_of_relation "multipolygon" or "boundary". Given that there are 852 relations, that's a lot of counties. A few are Nodes, which are single points: http://wiki.openstreetmap.org/wiki/Node and not very useful when solving the "map an arbitrary point to a tzid" unless there are a *LOT* of them, and some are Ways: http://wiki.openstreetmap.org/wiki/Way which appear to either be paths or small areas. I'd have to see the data to see what they are, however, and to see whether they're actually useful for the "map an arbitrary point to a tzid" problem (i.e., whether they can be used to define regions corresponding to tzdb zones).

On Sep 5, 2013, at 2:46 PM, Lester Caine <lester@lsces.co.uk> wrote:

Guy Harris wrote:

...

If all you have is longitude and latitude, you're going to have to do a geometric search.

Have you tried? Search for Connecticut http://www.openstreetmap.org/browse/relation/165794

Let's try a harder problem - let's search for Kentucky, instead: http://www.openstreetmap.org/browse/relation/161655 Note that it does *NOT* have a timezone= tag, and there's a good reason for that - more than one tzdb zone covers Kentucky. So let's look at Louisville: http://www.openstreetmap.org/browse/relation/1804307 Hmm. It doesn't have a timezone= tag, either. Let's try the county: http://www.openstreetmap.org/browse/relation/1804288 Nope, no timezone= tag there, either. Do you know of an administrative boundary that corresponds to a relation in OSM that *does* have a timezone= tag? (It should be timezone=America/Kentucky/Louisville, BTW. Paul, what particular administrative region of Kentucky would be the biggest one that should be tagged with timezone=America/Kentucky/Louisville? Or *is* there no such region?)

Now just like TZ, this is still work in progress to get the fine details, but a geometric search within that relation will be able at some point to respond with the timezone of this relation, unless a lower level relation provides it first.

And if you define a set of multipolygon or boundary relations that correspond to the actual borders of tzdb zones, you have only one level of relations to deal with, rather than walking down a hierarchy of those administrative boundaries that happen to be established in OSM (and possibly not finding one with a timezone= tag!)

The debate at the moment in relation to historic data is if start and end dates should control the visibility of data on the main map. It will on the historic version, but the base map just returns 'today'.

If someone has time to write a search routine that will access the API and return the timezone then we could have a location tool. All the data is there as usual it's just time which is needed :( Oh for that time machine ...

If by "all the data is there as usual" you mean that, for all points on the surface of the earth, they can be found within *some* relation that's tagged with the correct tzid, I'd need a citation for that claim.

On Sep 5, 2013, at 3:50 PM, Guy Harris <guy@alum.mit.edu> wrote:

So let's look at Louisville:

http://www.openstreetmap.org/browse/relation/1804307

Hmm. It doesn't have a timezone= tag, either.

Let's try the county:

http://www.openstreetmap.org/browse/relation/1804288

Nope, no timezone= tag there, either.

Not that there should be a difference: https://en.wikipedia.org/wiki/Louisville,_Kentucky "Since 2003, the city's borders have been coterminous with those of the county because of a city-county merger."

On Thu, Sep 5, 2013, at 21:50, Gwillim Law wrote:

My data are incomplete, but the time zone boundary in Kentucky has followed county lines since 1961, as far as I can tell. Before 1961 there were definitely some Kentucky counties that were divided between Eastern and Central. I've listed the results of my investigations at http://www.statoids.com/tus.html. Recent changes can usually be found described in dockets of the Department of Transportation.

The problem is that Louisville is one of those small areas that didn't consistently use DST in the 1970s, and it's unclear what locations around them followed their example.

Sent from my android device.

Guy Harris wrote:

...

If all you have is longitude and latitude, you're going to have to do a geometric search.

Have you tried? Search for Connecticut http://www.openstreetmap.org/browse/relation/165794

How did you find that? I tried (well, tried searching for new york state) and reached a dead end. Wifi here is usless ... when you use the left hand search when the results come up there is an option - I think call - view detail. This is a rapidly changing area of fuctionality currently.

I don't have the quick link on the phone ... Go to the osm wiki lookup timezone and the right hand info box should have a link to the tag statistics site. I'll push a link to my crib sheet later Sent from my android device. -----Original Message----- From: random832@fastmail.us To: lester@lsces.co.uk, tz@iana.org Sent: Fri, 06 Sep 2013 16:10 Subject: Re: [tz] tz dependency lat & lng On Fri, Sep 6, 2013, at 9:15, lester@lsces.co.uk wrote:

Wifi here is usless ... when you use the left hand search when the results come up there is an option - I think call - view detail. This is a rapidly changing area of fuctionality currently.

Okay, how do you get the list of objects that actually have a certain timezone value?

On Sep 6, 2013, at 9:52 AM, random832@fastmail.us wrote:

That's the problem I ran into the other day - the tag statistics site does not appear to offer a way to get a list of objects.

Exactly. They only offered aggregated statistics, not detailed queries, as far as I could tell.

http://global.mapit.mysociety.org/ Will be asking about adding timezone to the demo at the lecture shortly Sent from my android device so the quoting is crap!