Time Issues and Notes
From Spatial Data on the Web Working Group
Notes submitted by Chris Little after F2F3
- Accuracy versus Precision confusion is made worse by the only widely accepted temporal notation (ISO8601) which often conveys unjustified precision. E.g. A document update time of 2016-02-16T00:00:00Z is highly unlikely to imply that the document was updated at precisely midnight UTC.
- Numerous software artifacts conflate times and dates, and sometimes mistakenly assume dateTime objects can be used as a coordinate reference system.
- The current ISO8601:2004 standard states that the notation is meant to reflect an underlying Gregorian calendar, with leap seconds. Many occurrences of apparent ISO8601 timestamps ignore leap seconds and the tens of seconds difference between legal civil time and GPS time. Many people refer to the obsolete ISO8601:1988 as it is publicly available, whereas the current version has to be bought.
- There is an assumption that in this age, computer times, such as from a system log with IETF RFC 3339 timestamps, are comparable with sub-second precision. Even when NNTP protocols have been correctly implemented, this is difficult to achieve.
- Some countries only switched to the Gregorian calendar as late as the 1920s, so historical documents from between 1588 and 1923 may be several days adrift, depending on the country, giving inaccuracy in data extraction.
- There are many other calendar systems in use, and there is no agreed notation as to which is used, never mind agreed and accurate algorithms for their implementation. There is a rough categorisation of calendars into those that can be related directly to the Gregorian, because they are algorithmically based, and those that are fundamentally observation based, with a precision of minutes or even days. http://content.iospress.com/articles/semantic-web/sw187
- Overlying temporal ontologies may not allow valid reasoning if the underlying temporal 'microformats' are not accurate at the appropriate level.
- There is now an OGC Standards Working Group to produce standardised WKT (Well Known Text) to specify non-Gregorian calendars (e.g. no leap seconds, or no leap days, or 360 days/year, etc).
- There is a SDW WG proposal to minimally enhance the OWL Time vocabulary with non-Gregorian calendars.