Date precision and uncertainty

Precision and uncertainty are different

Precision is how fine the date is. A year, a month, a day or a second. Uncertainty is how sure the source is. EDTF keeps these apart. A year and month string such as 2004-06 is precise to the month, and adding ~ to it says the value is approximate ¹.

EDTF uses three qualifiers. The question mark means uncertain, the tilde means approximate, and the percent sign means uncertain as well as approximate. The example 2004-06-11% means the whole date is both uncertain and approximate ¹.

The three EDTF levels

The Library of Congress says Levels 1 and 2 specify features described in ISO 8601-2 ¹.

• Level 0 covers dates, date and time values, and intervals between dates. It specifies features of ISO 8601-1 ¹.
• Level 1 adds letter-prefixed years, seasons, qualifiers, unspecified rightmost digits and extended intervals ¹.
• Level 2 adds exponential years, significant digits, sub-year groupings, set notation and unspecified digits anywhere in a component ¹.

Partial, open and seasonal dates

• Unspecified digits. At Level 1, X may replace one or more rightmost digits to show the value of that digit is unspecified ¹.
• Open intervals. A double dot may be used when the start or end is not specified. 1985-04-12/.. is an interval from 12 April 1985 with an open end ¹.
• Unknown ends. A null string may be used for a start or end date that is unknown ¹.
• Seasons. Month values 21 to 24 stand for Spring, Summer, Autumn and Winter. 2001-21 means Spring 2001 ¹.
• Long years. A Y prefix marks a year of more than four digits, as in Y170000002 ¹.

Choices and estimates

At Level 2, square brackets wrap a list where one member is the answer, and curly brackets wrap a list where all members are included ¹. This lets a record say that an event happened on one of several dates, or on all of them.

Significant digits express an estimate inside a range. 1950S2 means some year between 1900 and 1999, estimated to be 1950 ¹.

Exact timestamps and their limits

RFC 3339, Date and Time on the Internet: Timestamps, was published in July 2002. It is a profile of ISO 8601 for the Gregorian calendar ². It says interoperability is best achieved with Coordinated Universal Time ².

Its form carries fractional seconds where needed, as in 1985-04-12T23:20:50.52Z ². An offset of -00:00 means the UTC time is known but the local offset is not ².

Machine time can carry its own margin of error. A trusted time-stamp token under RFC 3161 may include an accuracy field. Adding it to the stated time gives an upper bound and subtracting it gives a lower bound ³.

Examples side by side