Wrong Olson Timezone Offsets?

Question

Today I noticed that my script to convert and display Olson timezone IDs into Cities and GMT offsets was producing some very weird results: Argentina zones caught my attention because they were displaying a standard (non-DST) time offset of ±0000 in reference to GMT/UTC.

I examined my code and I found a small bug in my logic, however it wasn't related to the discrepancy, so I updated timezonedb via pecl to version 2012.8 but it's still returning the wrong offset...

Here is some code that returns the last 3 changes to the America/Argentina/San_Luis timezone:

$timezone = new DateTimeZone('America/Argentina/San_Luis');
$transitions = $timezone->getTransitions();

echo '<pre>';
print_r(array_slice($transitions, -3, null, true));
echo '</pre>';

And this is the output:

Array
(
    [59] => Array
        (
            [ts] => 1223784000
            [time] => 2008-10-12T04:00:00+0000
            [offset] => -10800
            [isdst] => 1
            [abbr] => WARST
        )

    [60] => Array
        (
            [ts] => 1236481200
            [time] => 2009-03-08T03:00:00+0000
            [offset] => -14400
            [isdst] => 
            [abbr] => WART
        )

    [61] => Array
        (
            [ts] => 1255233600
            [time] => 2009-10-11T04:00:00+0000
            [offset] => -10800
            [isdst] => 1
            [abbr] => WARST
        )
)

As you can see, the indexes 59 and 61 are DST offsets while the index 60 is a standard time offset.

However, if you check Time&Date the standard offset should be -10800 (3 hours) and not -14400:

Standard time zone: UTC/GMT -3 hours
No daylight saving time in 2012
Time zone abbreviation: ART - Argentina Time

In fact, even the abbr is wrong (should be ART since on 2009-10-10 DST was discontinued).

What's going wrong here? I'm pretty sure this is irrelevant but I'm running PHP 5.4.6 if it matters.

PS: I remember reading about some legal issues with the Olson database, might this be related?

---

UPDATE: I wrote a little script to compare the PHP non-DST offsets with this Wikipedia page:

function getStandardOffset($timezone)
{
    $timezone = new DateTimeZone($timezone);
    //$hemisphere = (ph()->Value($timezone->getLocation(), 'latitude') >= 0) ? 'north' : 'south';
    $transitions = array_reverse(array_slice($timezone->getTransitions(), -3, null, true), true);

    foreach ($transitions as $transition)
    {
        if ($transition['isdst'] == 1)
        {
            continue;
        }

        return sprintf('%+03d:%02u', $transition['offset'] / 3600, abs($transition['offset']) % 3600 / 60);
    }

    return false;
}

$diff = array();
$html = ph()->HTML->DOM(file_get_contents('http://en.wikipedia.org/wiki/List_of_tz_database_time_zones'));
$timezones = DateTimeZone::listIdentifiers();

foreach ($timezones as $timezone)
{
    $offset = str_replace('−', '-', ph()->HTML->DOM($html, sprintf('//td/a[contains(text(), "%s")]/../..', $timezone), '0.td.4.a'));

    if (strcmp($offset, getStandardOffset($timezone)) !== 0)
    {
        $diff[$timezone] = array
        (
            'php' => getStandardOffset($timezone),
            'wiki' => $offset,
        );
    }
}

echo '<pre>';
print_r($diff);
echo '</pre>';

The following time zones differ:

Array
(
    [America/Argentina/San_Luis] => Array
        (
            [php] => -04:00
            [wiki] => -03:00
        )

    [Antarctica/Casey] => Array
        (
            [php] => +08:00
            [wiki] => +11:00
        )

    [Antarctica/Davis] => Array
        (
            [php] => +07:00
            [wiki] => +05:00
        )
)

It's not a huge deal but I'm curious why this happens since I have the latest version of the Olson tzdb.

Answer 1

I'll take your first example here:

[60] => Array
    (
        [ts] => 1236481200
        [time] => 2009-03-08T03:00:00+0000
        [offset] => -14400
        [isdst] => 
        [abbr] => WART
    )

However, if you check Time&Date the standard offset should be -10800 (3 hours) and not -14400

From what I've read, this is not true; according to this article, around that time (March 3rd 2009) the timezone was changed to UTC-4 when there's no daylight saving and UTC-3 if there is. Though, the mentioned date in the article is March 14/15 and not March 8 ... I'm not sure what has caused that difference.