On Tue, 17 Mar 2009 21:55:44 +0000, john gilmore wrote: > >About negative leap seconds: they have always been envisaged as possible, and >provision was made ab initio for them. We came close to having one recently, >and they are not at all problematic: provided only that the time scale, e.g., >that embodied in STCKE values, associated with them is itself monotone >increasing, they need not be (as I made clear in an earlier post). > This time we agree. Mostly. You're using the modal sign convention for leap seconds. This is confused only slightly by IBM's following the convention, then correctly subtracting CVTLSO from the STCKE value, rather than adding.
But positive leap seconds are more likely problematic than negative leap seconds. I disagree still with an IBM employee who stated in these pages a few years ago that there is code in the OS such that should a "negative" leap second ever occur the system will be idled for that second to avoid the hazard of apparently decreasing UTC values. Rather, I suspect that is done routinely at "positive" leap seconds to spare the TIME macro the need to supply UTC values of 23:59:60.hh, and to spare applications the need to deal with them. Our site has abandoned leap seconds because of conflicting vendor behaviors. We run our TOD clock on UTC, and set CVTLSO to 0. The Sysplex Timer will adjust for any leap second over about half a day, smoothly. Do we "lose an hour" or "gain an hour" in Spring? I know how I adjusted my watch; I wouldn't describe it with either of those terms. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
