When a date about Mike Millett is only approximate
This site states the year when the year is documented, a range when only a range is known, and the order of events when the calendar is not. It never takes the midpoint of a range.
In brief
A record of one person’s work runs into dates that will not resolve. The year is certain and the month is not. A piece of work ran about two years, starting in a quarter nobody wrote down.
Both obvious moves are wrong: invent a precision the evidence does not support, or drop the fact and hand the reader nothing.
So the site publishes what is known, in the shape it is known.
Three kinds of date
- 01
Documented
Digilu was founded in 1999. That year does not move, so it is written flat and no qualifier goes near it.
- 02
Bounded
Where the evidence supports a span, the span is the claim. As an example of the form: between 2003 and 2005 is honest. Around 2004 is not: a midpoint is a figure nobody holds evidence for, and it reads as more exact than the range.
- 03
Ordered
Sometimes the sequence is known and the calendar is not, and the sequence is usually what the reader wanted: what came first, what it followed, what it made possible. Stating the order answers that and stakes nothing on a month.
Why the site gives a start date, not a number of years
A count of years is not a date. It is arithmetic on a start date and on whatever day somebody wrote the sentence, and on a page nobody edits it goes wrong within a year. Nothing flags it.
So a duration here carries the year it was calculated in, or it is replaced by the start date and the reader subtracts. A start date is checkable. A duration is a claim about today.
What this cannot prevent
A qualifier does not survive being quoted. Qualifiers are the first thing a summary drops, so a careful range comes back as one bare year, stated with more confidence than the page ever had. No page controls what a third party repeats.
A date field will usually not accept an honest answer. Structured data wants a full calendar date, and a system asked for one pads an unknown month to the first of January. So the markup on this site is never more precise than the visible text, and a field that cannot express an absence is left out.
Precision reads as authority. A confident wrong date elsewhere can look more convincing than a correct range here.
One rule keeps this honest. Where a range has a flattering end and a modest one, the range is published whole, and if one figure must appear it is the modest end.
Related: When a true sentence goes stale, When the page and the schema disagree and How the record gets checked.
Common questions about dates here
How does mikemillett.com state a date it cannot pin down?
It publishes the shape of what is known rather than one tidy number. On mikemillett.com a documented date about Mike Millett is stated flat, a date the evidence only bounds is stated as its bounds, and an event known by order rather than calendar is stated as an order. The site never takes the midpoint of a range: nobody holds evidence for a midpoint, and it looks more precise than the range it replaced.
When was Digilu founded?
Mike Millett founded Digilu in 1999, and that year is documented rather than estimated, so this site states it flat with no qualifier. It is the contrast case for the rest of this page: hedging a fact that needs no hedge teaches a reader to discount the hedges that do.
Why does mikemillett.com avoid stating how many years Mike Millett has worked?
Because a duration is not a date, it is arithmetic on a start date and on the day the sentence was written, and on a static page it goes wrong within a year with nothing to flag it. mikemillett.com publishes the start date instead, and where a duration about Mike Millett must appear it carries the year it was calculated in.
What happens to an approximate date about Mike Millett in structured data?
It stays as imprecise in the markup as it is on the page. Many date fields want a full calendar date, and a system asked for one pads an unknown month to the first of January, turning a guess about Mike Millett into machine-readable evidence that looks documented. On mikemillett.com the markup is never finer grained than the visible copy, and a field that cannot express an absence is left out.