Should the first in a queue be served last?
Everyone knows how a queue works. It's a line of people where the person at the front gets served first. But Danish researchers have recently made a shocking suggestion - that queuing on the basis of last-come-first-served may sometimes be more efficient.
The principle of first-come-first-served is simple, and it's fair. Who could possibly argue with it?
Well, one person is Prof Lars Peter Osterdal of the University of Southern Denmark.
"Queues, it's a wonderful example of a waste of time," he says.
"The problem with a regular queue where you serve first those who arrive first is that people tend to arrive too early."
Osterdal and his colleague, assistant professor Trine Tornoe Platz, studied situations where a service opens at a particular time and closes after every person has been served. Airlines that do not assign seats before boarding provide a good example.
Under the first-come-first-served system passengers arrive early and wait in line to get on to the plane because those who are first in line get the seat they want.
But the researchers experimented with different queuing systems, and when they told volunteer queuers that people would be selected from the queue and served at random, the average wait was reduced.
The best system, however, turned out to be last-in-first-out - or to put it another way, last-come-first-served - with the person who arrives last getting served first.
That system, Osterdal says, changes people's behaviour. They tend to arrive at staggered times, resulting in shorter queues.
"It would be more risky for people to arrive early because it could mean that you may not be lucky enough to be served immediately, so you would have to wait for a long time until all those who arrive after you have been served," he says.
"There will be some people trying their luck arriving early but on average people will arrive later and it means on average that everyone will be better off."
At the airport departure gate, people would be more likely to stay in a cafe for a while, or sit reading a book rather than rushing to be first in the queue.
I tried an experiment on my colleagues to see how they liked the last-come-first-served approach.
First I sent an email offering free chocolate fudge cake. Then, as people turned up, I made them form a queue and gave cake first to those at the back of the queue.
It's fair to say it wasn't a popular system among those who had queued longest. "Despicable," was the comment from the person at the front who had to wait longest for her slice.
The experiment was fun, but, of course, I missed out a key element of the last-come-first-served system. It's important that people about to enter the queue know how it works, so they can adjust their behaviour.
Osterdal does acknowledge, though, that the use of the last-come-first-served queue in the real world would be difficult in many cases.
"In practice it would be very hard to implement a principle of serving the last in most physical queues where people physically queue up, for example when boarding an airplane. It's simply too difficult to manage for practical reasons," he says.
It could be open to manipulation, with people leaving the queue and rejoining from the back in order to get served more quickly.
There's also the issue of fairness - or unfairness.
"As long as everyone knows it from the beginning, in a sense it's fair that everyone faces the same rules and the rules apply to everyone. But it's true that many people would object to serving the last because it will result in some people having to wait for a long time. In that sense it's not fair and probably it's against many people's intuition about fairness in queues," Osterdal says.
Abandoning the first-come-first-served principle would be easier in internet or phone queues though, Osterdal points out.
People kept waiting a long time may not be happy, but at least they would not see people who arrived in the queue after them getting served first.
Subscribe to the BBC News Magazine's email newsletter to get articles sent to your inbox.