Re: User Petition on Standards to Netscape and Microsoft

["Louis A. Mamakos" <louie at UU.NET>]

> In my opinion, GPS's lack of leap seconds makes it vastly preferable
> to UTC for use as a basic, monotonically increasing internal time
> scale in computer protocols and operating systems. Leap seconds
> introduce any number of problems; just consider the transient effects
> on NTP every time a leap second occurs. Dave Mills may consider them
> really fun to watch, but I consider it a basic design flaw of NTP that
> such transients occur in the first place.

This isn't so much a problem with NTP, but with the implementation of the
local clock model.  If ctime(), et. al. presumed a monotonically increasing
timescale and consulted a database of leap seconds, you wouldn't introduce
a discontinuity.  That the time scale that the UNIX kernel uses "includes"
leap seconds is the botch.

Of course, you've now made the conversion between this time scale and 
a particular instance of wall-clock time much more complicated than adding 
a fixed offset.  Grief is conserved.

louie