Dijkstra’s zero-indexing argument starts with a list that contains nothing
A colleague’s accusation of pedantry prompted the 1982 note; the useful payoff is a way to check range boundaries without an extra adjustment.

Dijkstra defended counting from zero by asking what happens when there is nothing left to count. The empty list finally got its moment, which seems only fair after doing absolutely nothing to cause the argument.
His August 11, 1982 note, preserved by the University of Texas archive, starts with the sequence from 2 through 12. He compares four ways to write its bounds. The preferred form includes the lower endpoint and excludes the upper: 2 ≤ i < 13.
The length then comes straight from subtraction: 13 minus 2 equals 11. Adjacent ranges meet at the same boundary. The values 2 through 12 end where a following range begins, without making either side argue over who gets the shared number. Off-by-one errors have plenty of other job opportunities.
Dijkstra uses the empty sequence to choose which endpoint to include. Keeping the lower endpoint included avoids an unnatural lower bound for a sequence starting at the smallest natural number. Excluding the upper lets that sequence shrink to nothing without forcing an unnatural upper bound either.
He then applies the convention to a sequence of length N. Starting at zero gives 0 ≤ i < N, and an element’s index counts how many elements precede it. His note also cites experience with Mesa at Xerox PARC as supporting evidence for the interval choice.
Try a small example on paper: a range from 0 up to, but excluding, 3 contains 0, 1 and 2. Splitting it at 1 produces a first part from 0 to 1 and a second from 1 to 3. The boundary can be copied directly between the parts, and their lengths still add to three.
Shrink the full example to an end of 0. The range from 0 to 0 contains nothing, and subtraction still produces the correct length. That gives the empty case the same arithmetic as the populated case, rather than requiring a little explanatory asterisk every time the list runs out.
The provocation behind the note was social as well as mathematical. Dijkstra says a mathematical colleague accused younger computing scientists of “pedantry” for starting at zero. That is his account of why he wrote it, rather than a survey of every programmer’s motives.
A convention is easier to defend when its awkward cases behave cleanly. It can still be unfamiliar to somebody who expects the first item to be numbered one. The useful question for a particular API is which convention it actually promises and whether its callers share that expectation; this one page cannot settle what every existing system ought to change.
Zero brought an empty list.