Leap seconds are scheduled by the International Earth Rotation Service (IERS) whenever the difference between true earth rotation and the UTC time scale reaches a certain limit. Whenever a leap second has been scheduled by the IERS, a warning must be disseminated to time keeping devices so that clocks become aware of the scheduled leap second early enough to be able to handle the leap second properly.
There are different ways to propagate leap second warnings using different timing signals, protocols, etc. For example, the GPS satellites transmit a specific point in time when a leap second is to be inserted or deleted, but other timing signals may just provide a leap second warning flag which is set during a certain interval before the leap event, where the warning interval depends on the specification of the protocol.
Also, there are different implementations how leap seconds are handled, which especially affect the sequence of timestamps across the leap second event. The clock can be stepped at the beginning or end of the leap second, can be slowed down or even stopped during a leap second insertion, or time can be slewed across a leap second. This makes it difficult to compare time stamps which have been taken on different systems during a leap second.
Last, but not least, there are implementations of time keeping code which don't always work correctly, e.g. invalid leap second warnings are generated, leap seconds are not handled at all, or severe bugs occur due to side effects of the leap second handling.
Speakers: Martin Burnicki