quality of historical zone information

Gwillim Law gwil at mindspring.com
Mon Jan 29 18:20:18 UTC 2001


Alois Treindl wrote:
> Defining more zones will not be enough.
> it must be possible to associate every populated place with the right
> zone.

That's a requirement for my work, too.  I have been working in the field of
logistics.  To track shipments, I need to be able to convert from local time
to UTC, based on location.  That means I have to know the historical
sequence of UTC offsets for each possible location.  The difference between
my application and astrology is that I'm concerned with a much narrower
range of dates:  typically from today minus three months to today plus three
months, although for archival purposes I may want to look back several
years.

If the membership of the tz list, or any other group, wants to undertake to
create and maintain a more complete tz archive that would do justice to the
geographical aspect, I have a few principles to suggest.

Too much work is involved to expect any one person, especially a volunteer,
to do it all.  There's a natural way to subdivide it:  by countries.  In
general, time zones are determined by legislation which applies throughout a
jurisdiction, almost always a whole country.  In many cases, a volunteer
could take responsibility for several countries or a whole region.  On the
other hand, a country with a very complex time zone history, such as the
United States, might be divided up among several volunteers.  Ideally, each
volunteer would be acquainted with the languages and culture of the region
he or she is responsible for.  Uniform guidelines should be drawn up,
covering things like database design, data authentication, and update
notification.

Database design:  The current tz archive has an implicit database design,
which is implemented in various Posix-compliant systems.  This existing
design is inadequate for some purposes.  For example, Brazil is divided into
areas that observe four different standard times.  The information about
which state is in each area is hidden in the comments.  If a new database
design allows the geographical component to be represented, Posix systems
should be supported as a legacy.  This might be done by taking a slice or
subset of the data.  Other systems might choose to take a different slice,
in order to tailor the data to their specific requirements and save on
storage.  A logistics application might select only post-1996 data, having
no need to refer further back than that.

Data authentication:  the data in the existing tz archive were acquired from
several sources.  They're cited in the comments.  They include Shanks,
Whitman, IATA, and personal communications from a number of Time Zone
Caballeros.  None of these are really primary sources.  Primary sources
include laws, acts, and decrees reported in the Congressional Record,
Hansard, etc., and newspaper reports of actual time zone implementation.  We
do have some primary sources, but they don't cover the whole world.  When
several sources disagree (and we don't have a primary source), someone like
Paul Eggert decides which is the most reliable, and explain his decision in
the comments.  As new sources become available, some of the old decisions
may have to be reconsidered.  What we need is a file containing every
available document that has clues to historical time zone data.  This file
could be partitioned by country or region, so that each time zone volunteer
saves only those documents pertaining to his or her specific area.
Guidelines should cover criteria for selecting the most reliable information
from a set of documents, as well as the maintenance, back-up, and public
availability of the source material.

Update notification:  Some updates to the expanded tz archive would be to
correct old errors.  Others would be to extend time zone rules further into
the future.  Currently, all updates to the tz archive affect all users, and
are reported to the tz mailing list.  Would this still be sufficient?  It
would be a great help to maintain a Web site that displays all the tz
archive data in a format easily understood by casual users.  For one thing,
more people would be able to find mistakes in the data and send in
corrections.  Perhaps Web pages could be automatically generated by applying
a transformation to the archive.

The work I've described above would be a valuable service to the Internet
community.  Is there some way to finance it?  For example, could certain
beneficiaries be persuaded to subsidize it?  Could advertising be sold on
the Web site?  If so, the volunteers could be paid, and the archive could be
more effectively improved.

Yours,    Gwillim Law





More information about the tz mailing list