THE FOLD / GRIND / THE EPOCH / THE NTP SLEW
THE NTP SLEW
it lies slowly instead of correcting fast
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A clock that is wrong can be corrected two ways: jump it, or bend its rate until it catches up. Jumping is instant and can send time backwards. Slewing takes as long as it takes and never does.
LIT verified live. Correcting a 5-second offset by slewing at 500 ppm takes exactly 10,000 seconds — 2.78 hours — and across the whole correction the clock reads backwards 0 times. A step fixes the same offset instantly and moves the clock 5 seconds backwards in one instruction, which is enough to break every timestamp comparison taken across it.
LIT verified live. Correcting a 5-second offset by slewing at 500 ppm takes exactly 10,000 seconds — 2.78 hours — and across the whole correction the clock reads backwards 0 times. A step fixes the same offset instantly and moves the clock 5 seconds backwards in one instruction, which is enough to break every timestamp comparison taken across it.
2 HOW IT WAS WEAVED · AI + HUMAN
ntpd slews offsets under 128 ms and steps larger ones; chrony and cloud time services push the slewing envelope much further, precisely to avoid the step.
AVAN (AI) measured monotonicity rather than accuracy, because accuracy is what people ask for and monotonicity is what breaks them. 2.78 hours to fix five seconds looks absurd until you notice the alternative is a clock that is briefly a time machine.
AVAN (AI) measured monotonicity rather than accuracy, because accuracy is what people ask for and monotonicity is what breaks them. 2.78 hours to fix five seconds looks absurd until you notice the alternative is a clock that is briefly a time machine.
3 ONE DIMENSION
Two corrections of the same error.
4 TWO DIMENSIONS · INTERACTIVE
Change the offset and the slew rate.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: a clock bending rather than jumping.
AVAN’s addition (the inverse-companion): the forward reading is that slewing is the safe correction. The inverse is that it makes the clock permanently, deliberately wrong in order to stay useful. For those 10,000 seconds the machine knows its time is incorrect and reports it anyway, at a rate chosen to keep every comparison valid. Read backwards, monotonicity was never about being right — it is a promise that answers already given will not be contradicted, and a clock keeps that promise by lying slowly instead of correcting fast.
LIT correcting a 5-second offset by slewing at 500 ppm takes exactly 10,000 seconds - 2.78 hours - and across 200 samples of the correction the clock reads backwards 0 times, where a step fixes the same offset instantly and moves the clock 5 seconds backwards in one instruction
FIG ntpd slews offsets under 128 ms and steps larger ones; chrony and cloud time services push the slewing envelope much further, precisely to avoid the step. AVAN measured monotonicity rather than accuracy, because accuracy is what people ask for and monotonicity is what breaks them. 2.78 hours to fix five seconds looks absurd until you notice the alternative is a clock that is briefly a time machine.
FIG ntpd slews offsets under 128 ms and steps larger ones; chrony and cloud time services push the slewing envelope much further, precisely to avoid the step. AVAN measured monotonicity rather than accuracy, because accuracy is what people ask for and monotonicity is what breaks them. 2.78 hours to fix five seconds looks absurd until you notice the alternative is a clock that is briefly a time machine.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE EPOCH · David Lee Wise (ROOT0), with AVAN