[LEAPSECS] LEAPSECS Digest, Vol 71, Issue 4

[Warner Losh]

On Sep 4, 2012, at 11:25 AM, Finkleman, Dave wrote:

> With respect to Warner's statement about how robust leap seconds are,

> "robust" with respect to what application?   IMO, the scheme is robust

> for those whose missions depend on Earth sync.   It is not robust for

> those with at best casual interest and whose missions are hardly

> affected by leap second implementation anomalies.


They are fragile in comparison to leap days.  Everybody knows leap days, and agrees on them for the next thousand years or so.  They are engrained in the culture and in the technology.  They are robust.

Leap seconds are, by comparison fragile.  They insinuate themselves into many nooks and crannies and have proven difficult to get right.  Some of that is the attitude that one second errors aren't worth spending the engineering resources on.  Part of it is ignorance of problem.  Part of it is their irregularity and non-set schedule.  We had many years where there wasn't a leap second, so people got used to it not being an issue.  Plus, the 6-month horizon doesn't fit the annual budget process well, leading to surprises in people's operating budget, or to black eyes for folks that didn't test their systems.  That is why I say they are fragile: it isn't just outliers that need to care, and it isn't just outliers that get them wrong.

The ntp announcement thing is just one more brick in this rather large wall.

Warner