Skip to content

Home

Jump to bottom
Chris Little edited this page Dec 1, 2021 · 32 revisions

Welcome to the Temporal-Abstract-Spec wiki

First cut at structuring resources to input into the Abstract spec

  1. Minutes of the OGC Temporal Domain Working Group

  2. Existing standards

  3. Restrictive profiles of existing standards

  4. Other useful Documents

  5. Stakeholders in temporal infrastructure

    ISO

    OGC Naming Authority

    BIPM

    IERS

  6. Examples of good and bad practice

  7. Examples or catalogue of temporal systems

  8. Draft text

Quid est ergo tempus?

Si nemo ex me quaerat, scio.

Si quaerenti explicare velim, nescio.

  Saint Augustine, Confessions, XI

What then is time?

If no one asks me, I know.

If I wish to explain to him who asks, I do not know.

Temporal Regimes

This conceptual model proposes four regimes:

  1. Event Regime - No clocks, logic only
  2. Clock Regime - Clocks, ticks, integer arithmetic only, no –ve times
  3. Coordinate Regime - CRSs, number line to interpolate between ticks, real arithmetic, extrapolate before zero/datum/epoch
  4. Calendar Regime - Calendars, abnormal arithmetic, earth sun and moon rotations, months, weeks

1. Event Regime

Set of EVENTS ordered in time (ta, tb, tc, … tn), that may be: Finite or Countably infinite (like the integers)

No clocks

Simple Logic Operators defined to determine if 2 times are: The same, or one earlier, the other later (maybe an earliest and latest times?)

2 times define Relation(ta,tb), ta<tb, but NOT duration (tb-ta)

Any 3 or 4 events could allow logic (Allen 1983) such as:

 tb is in (ta,tc), ta<tb<tc

 (ta,tc) overlaps (tb,td), ta<tb<tc<td, 

 (ta,td) contains (tb,tc ), ta<tb<tc<td, 

 (ta,tb) disjoint from (tc,td), ta<tb<tc<td,

No other times exist or can be interpolated

Similar to RCC8 spatial operators

Not sure about continuum. Probably OK. Operations are the distinguishing features.

Examples: Geology, Archaeology, Ice and sediment cores, King Lists and Genealogies

Could be multiple sets, each with own ordering, may or may not be relations between events across sets

ISO8601 notation 2014-10-07T10:00 is a BAD notation for this regime. Notation should just be labels E.g. “Jurassic”, “65My ago”, “Reign of Tiglathpileser III”, "1.562m from top of core" (?? not sure about ice cores? Tree rings are definitely Regime 2, clocks)

2. Clock Regime

“Clock” defined as any regularly repeating physical event that can be counted (or perhaps measured?)

Countably infinite set of ordered time INSTANTS.

Fixed precision, determined by interval between instants

No intermediate times can be calculated between ticks

Similar set of operations: Same, Earlier/later, Operations for Instants and Intervals

 tb is in (ta,tc), ta<tb<tc

 (ta,tc) overlaps (tb,td), ta<tb<tc<td, 

 (ta,td) contains (tb,tc), ta<tb<tc<td, 

 (ta,tb) disjoint from (tc,td), ta<tb<tc<td,

Can now calculate duration (tb-ta) as metric defined

Could be an earliest or latest, or an epoch (datum time)

Examples:

  • TAI International Atomic Time, which cannot be used to time events to femtosecond precision, or cannot be used to time events prior to Epoch!
  • Pendulum swings
  • Heart-beat
  • Vibration of a piezo-electric crystal
  • Rotation of earth around sun

Notation implying higher accuracy than ticks not appropriate. E.g. Julian days labelled 2014-10-07T10:00:00.0+01:00

3. Coordinate Regime

Precision defined with countably infinite set of ‘ticks’

Assume normal mathematical interpolation between ticks

Epoch defined, perhaps with practical earliest/latest times

Assume mathematical extrapolation before epoch or into the future: +/-

Logical Operations and calculations well defined

Other ‘UoM’ can be defined but must be totally ‘regular’ E.g: 1 hour = 60 mins, 1 day=24 hours, 1 year = 360 days E.g: milliseconds, kiloseconds

Examples: Unix milliseconds, Julian Days, Julian Years, etc

Epoch may be ill-defined (start of reign of Ashburnipal III) so CRS is relative, not absolute – see Regime 0.

Epoch should be defined in terms of TAI, UTC, etc

Regimes 1 & 2 blend

Think carefully about timestamp notation, such as ISO8601: 2014-10-07T11:00:00.0 actually implies LOCAL time, which could be solar! 2014-10-07T10:00:00.0Z implies UTC, and therefore leap seconds

Astronomers epoch is usually at midday, everyone else prefers midnight, of specified day

4. Calendar Regime

Anything requiring an algorithm beyond normal arithmetic

E.g: Years CE and BCE (AD & BC). There is no year 0BCE or 0CE, so ‘normal arithmetic’ gives unexpected results E.g. UTC Gregorian, Mayan, Jewish, etc

Should have an Epoch

May have earliest and latest defined times, or times when algorithm invalid

Usually approximates a CRS to astronomical events

Algorithmic rather than observed calendar

A CRS is not a Calendar, a Calendar is not a CRS!

Other topics

Notation is not a Timescale, CRS or Calendar

ISO 8601:2004: 2014-10-07T10:20:00.0

Is 0000-01-01T00:00:00.0 valid?

2013-07-01T00:00:00 minus 2013-06-30T23:59:00 = ?

59 or 60 or 61 seconds?

2012-07-01T00:00:00Z minus 2012-06-30T23:59:00Z = ?

59 or 60 or 61 seconds?

Do not assume notation implies arithmetic, CRS or calendar!

What notation applicable for each regime?

Answers: 2012-07-01T00:00:00 minus 2012-06-30T23:59:00 = 60 possibly, if local time is UTC

2012-07-01T00:00:00Z minus 2012-06-30T23:59:00Z = 61 precisely because ‘Z’ implies UTC, and leap seconds only introduced at midnight UTC

Mixed Regimes

Astronomical times, local solar time, sidereal time, relativistic, helio-spatial, accountancy, etc Requires observation of moon, sun or stars Could be several regimes: Observation based calendars Sidereal time Local Solar time, Mean Tropical Year Space weather time on Sun Relativistic Etc Plenty of realistic detailed use cases Accountancy? Weeks and months! Plenty of calendars in this regime

Time Scales and Calendars

All defined timescales and calendars, as they are based on physical devices, have a temporal scope; That is, a time inerval for which they are valid. Outside of the this interval, people may still choose to use that temporal reference system, but with indeterminate accuracy and precision, and no method of determining them.

registering temporal CRSs and Calendars

Simple Temporal CRS form for registration Clearly specified and determined datum (epoch) May be absolute (E.g. specified in UTC or TAI) Or relative (start of ice core, start of Tiglathpileser III’s reign) Well defined and named unit of duration Well defined directions (+ and -) Normal arithmetic No missing or extra years, seconds, etc There is a value of zero at the datum There may be ‘earliest’ or ‘latest’ practical values Sensible CRS name Passes OGC-NA criteria URI scheme Has convincing use case to be separate from existing CRSs

Some existing Coordinate reference systems in OGC NA http://www.opengis.net/def/crs/OGC/0/AnsiDate http://www.opengis.net/def/crs/OGC/0/ChronometricGeologicTime http://www.opengis.net/def/crs/OGC/0/JulianDate http://www.opengis.net/def/crs/OGC/0/TruncatedJulianDate http://www.opengis.net/def/crs/OGC/0/UnixTime Coordinate system axes: http://www.opengis.net/def/axis/OGC/0/days http://www.opengis.net/def/axis/OGC/0/mya http://www.opengis.net/def/axis/OGC/0/seconds Datums: http://www.opengis.net/def/datum/OGC/0/AnsiDateDatum http://www.opengis.net/def/datum/OGC/0/JulianDateDatum http://www.opengis.net/def/datum/OGC/0/TruncatedJulianDateDatum http://www.opengis.net/def/datum/OGC/0/UnixTimeDatum http://www.opengis.net/def/datum/OGC/0/YearZeroDatum

ANSI Date days elapsed from 24-Nov-4714 BC (12h00 UTC), proleptic Gregorian calendar.

Clone this wiki locally