One additional second is all one needs, no? Two leap seconds are never inserted adjacent to each other. So POSIX's 'struct timespec' suffices to >represent leap seconds, and there's no need for a new data type here, just new operations on the type. You could start by adding a new clock to >clock_gettime, a clock that respects leap seconds.

There has never been more than one leap second added in any one instance, but as far as I know, nothing precludes adding more than one at any one time. But, I daresay there would be a really significant event that would cause it, such as an asteroid strike that affect the day :) Nathan