DTV 1.0b2 FTF Avatar
  1. OMG Issue

DTV — basic time coordinate concepts are badly described

  • Key: DTV-40
  • Legacy Issue Number: 17428
  • Status: closed  
  • Source: Thematix Partners LLC ( Edward Barkmeyer)
  • Summary:

    Figure 12.1 shows 'time coordinate' as a specialization of SBVR 'representation', which is said to be the relationship between an expression and the meaning it denotes. No UML diagram shows 'time expression' (from 12.1.2), which one would expect to be the corresponding specialization of 'expression'.

    In 12.1, the definition of 'time coordinate' is: "representation of a time point by an atomic time coordinate or compound time coordinate that indicates the time point". This is both circular and incorrect. A time coordinate cannot represent a time point by a subtype of itself, it must represent a time point by an expression. The definition should read: "representation of a time point by a time expression" (to exclude representation by a definite description).

    One would expect a 'time expression' to be "the expression of a time coordinate" (from SBVR 'representation has expression', mislabeled 'expression uses representation' in the diagram), but it isn't. It is said to be an expression involving an index (integer) and a time scale, which means it is not the expression of a compound time coordinate.

    In 12.1.2, the definition of 'atomic time coordinate indicates time point' requires the time expression to contain an index, which means that the "February" example is invalid. The expression (string) "February" includes neither an index nor a time scale. The time coordinate, as a 'representation', is the association of "February" with the time point that is month of year 2. The atomic time coordinate thus acquires the time scale and index properties from the time point it indicates. But the expression "February" does not have those properties. "February" is associated with that time point via 'concept has designation'. If DTV is to explain how a time expression is associated with a time point, it has to distinguish the properties of the expression from the properties of the association (that SBVR calls a 'representation').

    A 'simple time expression' is either the expression of an 'index', which would be some expression of an integer, OR a 'signifier' for a time point (from SBVR 'concept has designation' and 'designation has signifier (expression)'). The signifier could be a given name, like "March", or a constructed term involving the scale and the index, like "Gregorian month 3". (What 12.1.2 describes is only the last case.)
    A 'compound time expression' is some syntax that combines two or more simple time expressions.

    In SBVR style, then, an atomic time coordinate is a time coordinate whose expression is a simple time expression, and a compound time coordinate is a time coordinate whose expression is a compound time expression.

    And the rules for determining what a simple time expression indicates, i.e., how the link that is the atomic time coordinate is constructed, depend on the nature of the time expression. In particular, the context of a compound time expression may make the intent of an index expression clear, as in "3/31/2012", where the "3" is only recognized as a month of year index because of its placement.

  • Reported: DTV 1.0b1 — Thu, 14 Jun 2012 04:00 GMT
  • Disposition: Resolved — DTV 1.0b2
  • Disposition Summary:

    The definition of ‘time coordinate’ is simplified to say just that it is a representation of a time point. To support this:

    • ‘time expression’ and related concepts are deleted
    • ‘atomic time coordinate’ is a time coordinate that has an expression that is either the name of a time point or a time point kind and an index
    • ‘compound time coordinate’ is redefined as a “time coordinate that combines more than one atomic time coordinate”
    • ‘time coordinate indicates time point’ is defined in terms of SBVR's 'representation represents meaning'
    • 'compound time coordinate indicates time point' is redefined in terms of the existing verb concept 'compound time coordinate combines atomic time point'

    Other miscellaneous changes:
    • Several new reference schemes are added to 'time point'
    • The text at the start of clause 10.5.1 is instead made the introductory text for all of 10.5.
    • References to the clauses where time coordinates are defined are fixed.
    • The definitions of ‘absolute time coordinate’ and ‘relative time coordinate’ are corrected
    • Missing SBVR concepts are added to clause 4
    • Delete 'atomic time coordinate has index', which was added by the resolution of issue 17426. This concept is no longer required.

    The issue summary made the comment that “the context of a compound time expression may make the intent of an index expression clear, as in "3/31/2012", where the "3" is only recognized as a month of year index because of its placement.” This specification does not describe the internal structure of a time coordinate, NOT the external representation of one. The interpretation of a date such as “3/31/2012” is a function of a tool, not of this specification.

    Note: this issue is dependent upon 16951, which defines 'time point kind'.

  • Updated: Fri, 6 Mar 2015 20:58 GMT