- From: Little, Chris <chris.little@metoffice.gov.uk>
- Date: Wed, 8 Jun 2016 12:22:25 +0000
- To: "Simon.Cox@csiro.au" <Simon.Cox@csiro.au>, SDW WG Public List <public-sdw-wg@w3.org>
- CC: "eparsons@google.com" <eparsons@google.com>, "kerry.taylor@anu.edu.au" <kerry.taylor@anu.edu.au>, "phila@w3.org" <phila@w3.org>
- Message-ID: <3DAD8A5A545D7644A066C4F2E82072883E203864@EXXCMPD1DAG4.cmpd1.metoffice.gov.uk>
Dear Simon and Colleagues, 0. My main concern about Temporal Reference Systems is the major splits between event driven (archaeological layers, geological periods, King Lists, …), simple clocks with countable timescales (Atomic time, pendulums, heart beats, rotations of astronomical bodies, …) and calendars (Gregorian, Hindu, Jewish, Muslim, Yoruba, …) 1. I am not completely clear about the scope of ‘ordinal systems’. Finite ordered sets of events, as in geological periods, tree rings for a specific tree, etc, are clear, but is a set of clock ticks, possibly of infinite number, considered ‘ordinal’ or not? Maybe it does not matter. 2. All of the three major categories above could support ‘interpolation’ and ‘extrapolation’ using some kind of measure. For example: discrete, countable, atomic time can be interpolated at the femto-second level, even if not very meaningfully, and extrapolated before and after the existence of a physical clock. Similarly for other clock based timescales. This gives us the CRS-like TRS. I suspect measures on most ordinal systems are not very meaningful and do not add value above the Allen operator based reasoning about the order of events. 3. Calendars, down to the resolution of days, are complicated algorithms, with multiple measures/units/durations/periodicities and concatenating them with clock based systems just makes them worse. I suppose that the above is my first stab at a fundamental/higher level temporal ontology? Added to the above I suppose I ought to add that: 4. Basically ISO8601 is only a notation for expressing a very limited subset of the possible underlying semantics. Chris Little FWIW - here is the text from ISO 19108 introducing the models for temporal reference systems, courtesy of Simon Cox: 5.3 Temporal reference systems 5.3.1 Types of temporal reference systems A value in the time domain is a temporal position measured relative to a temporal reference system. ISO 8601 specifies the use of the Gregorian Calendar and 24-hour local or Coordinated Universal Time (UTC) for information interchange. This shall be the primary temporal reference system for use with geographic information. A different temporal reference system may be appropriate for some applications of geographic information. In this case, the feature catalogue or the metadata associated with an application schema or data set shall include either a citation for a document that describes that temporal reference system, or a description of that system. When more than one temporal reference system is used in a single feature catalogue, application schema, or data set, the definition of each temporal characteristic shall identify the temporal reference system that is used. This subclause describes a conceptual schema that shall be used as the basis for such descriptions. Annex C defines metadata elements derived from this schema that shall be used in such a description. The Temporal reference system package includes three common types of temporal reference systems: calendars (used with clocks for greater resolution), temporal coordinate systems, and ordinal temporal reference systems (see Figure 7). [UML diagram] Figure 7 — Temporal reference systems The class TM_ReferenceSystem shall provide the following attributes. a) name: RS_Identifier shall provide a name that uniquely identifies the temporal reference system. The data type RS_Identifier is defined in ISO 19111. b) DomainOfValidity: EX_Extent shall identify the space and time within which the TM_ReferenceSystem is applicable. The data type EX_Extent is specified in ISO/TS 19103. It permits a description of both spatial and temporal extent. This attribute shall be used whenever an application schema includes TM_TemporalPositions referenced to a TM_ReferenceSystem which has a valid extent that is less than the extent of a data set containing such values. The following three subclauses describe the schemas for the three reference system types. 5.3.2 Calendars and clocks 5.3.2.1 Introduction Calendars and clocks are both based on interval scales. A calendar is a discrete temporal reference system that provides a basis for defining temporal position to a resolution of one day. A clock provides a basis for defining temporal position within a day. A clock must be used with a calendar in order to provide a complete description of a temporal position within a specific day. Figure 8 provides the details of the classes TM_Calendar and TM_Clock. Calendars have a variety of complex internal structures. This schema defines a simple external calendar interface. Annex D provides a more detailed description of the internal structure of calendars. Every calendar provides a set of rules for composing a calendar date from a set of elements such as year, month, and day. In every calendar, years are numbered relative to the date of a reference event that defines a calendar era. A single calendar may reference more than one calendar era (See D.3.1 and D.3.2 for examples). ... 5.3.3 Temporal coordinate systems Specifying temporal position in terms of calendar date and time of day complicates the computation of distances between points and the functional description of temporal operations. A temporal coordinate system may be used to support applications of this kind. A temporal coordinate system shall be based on a continuous interval scale defined in terms of a single time interval. [UML diagram] Figure 9 — Temporal coordinate system 5.3.4 Ordinal temporal reference systems In a number of applications of geographic information — geology and archaeology, for example — relative position in time is known more precisely than duration. The order of events in time can be well established, but the magnitude of the intervals between them can not be accurately determined. An ordinal temporal reference system is appropriate in such cases. An ordinal temporal reference system is based on an ordinal scale. In its simplest form, an ordinal temporal reference system is an ordered series of events. Generally, a specific series of events is associated with a single location. Temporal relationships between different locations can be determined only to the degree that events at one location can be correlated with events at other locations on the basis of non-temporal characteristics of the events. Such correlation can be used to develop a more broadly based temporal reference system defined in terms of periods within which similar events have occurred. The term ordinal era is used in this standard to refer to such a period. An ordinal temporal reference system consists of a set of ordinal eras (see Figure 10). Ordinal reference systems are often hierarchically structured such that an ordinal era at a given level of the hierarchy includes a sequence of coterminous shorter ordinal eras. [UML diagram] Figure 10 — Ordinal temporal reference system From: Simon.Cox@csiro.au<mailto:Simon.Cox@csiro.au> [mailto:Simon.Cox@csiro.au] Sent: Thursday, May 19, 2016 5:46 AM To: Little, Chris Cc: eparsons@google.com<mailto:eparsons@google.com>; kerry.taylor@anu.edu.au<mailto:kerry.taylor@anu.edu.au>; phila@w3.org<mailto:phila@w3.org> Subject: ISSUE-66: Should we define a vocabulary for describing temporal reference systems - Spatial Data on the Web Working Group Tracker Importance: High Chris – one for you to think about … https://www.w3.org/2015/spatial/track/issues/66 Simon
Received on Wednesday, 8 June 2016 12:22:57 UTC