THE FOLD / LOOT / THE INVENTORY / THE FRAME OF REFERENCE
THE FRAME OF REFERENCE
it compresses nothing and merely stops repeating yourself
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A column of timestamps looks like large numbers and is really a small range with a big offset. Store the minimum once and the distances from it, and the values shrink to fit the spread rather than the magnitude.
LIT verified live. 100,000 timestamps around 1,700,000,000 spanning 999. The deltas need 10 bits each rather than 32 — 125,004 bytes against 400,000, a factor of 3.2. Every value reconstructs exactly: 0 errors across all 100,000, because the transform is subtraction and nothing is approximated.
LIT verified live. 100,000 timestamps around 1,700,000,000 spanning 999. The deltas need 10 bits each rather than 32 — 125,004 bytes against 400,000, a factor of 3.2. Every value reconstructs exactly: 0 errors across all 100,000, because the transform is subtraction and nothing is approximated.
2 HOW IT WAS WEAVED · AI + HUMAN
Frame-of-reference encoding is standard in column stores and time-series databases, usually stacked with bit-packing on top of it.
AVAN (AI) verified exact reconstruction rather than only measuring the ratio, because a compression figure means nothing without it. The number that carries the idea is 10 bits: the data was never 32 bits wide, it was 10 bits of information wearing a 32-bit costume, and the encoding does not compress anything — it stops storing a constant a hundred thousand times.
AVAN (AI) verified exact reconstruction rather than only measuring the ratio, because a compression figure means nothing without it. The number that carries the idea is 10 bits: the data was never 32 bits wide, it was 10 bits of information wearing a 32-bit costume, and the encoding does not compress anything — it stops storing a constant a hundred thousand times.
3 ONE DIMENSION
The values, and the same values minus their minimum.
4 TWO DIMENSIONS · INTERACTIVE
Widen the spread and watch the saving disappear.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: one offset, and a hundred thousand small numbers.
AVAN’s addition (the inverse-companion): the forward reading is that frame-of-reference compresses a column dramatically. The inverse is that it compresses nothing and merely stops repeating yourself. The 22 high bits were identical in every row; they were never data, they were a fact about the column stored once per value. Read backwards, most impressive compression ratios are measurements of how much redundancy the format introduced in the first place, and the honest figure is not 3.2× but 10 bits — the amount that was ever there.
LIT 100,000 timestamps around 1,700,000,000 spanning 999 need 10 bits per delta rather than 32 - 125,004 bytes against 400,000, a factor of 3.2 - and every value reconstructs exactly with 0 errors across all 100,000, because the transform is subtraction and nothing is approximated
FIG Frame-of-reference encoding is standard in column stores and time-series databases, usually stacked with bit-packing on top. AVAN verified exact reconstruction rather than only measuring the ratio, because a compression figure means nothing without it. The number that carries the idea is 10 bits: the data was never 32 bits wide, it was 10 bits of information wearing a 32-bit costume, and the encoding stops storing a constant a hundred thousand times.
FIG Frame-of-reference encoding is standard in column stores and time-series databases, usually stacked with bit-packing on top. AVAN verified exact reconstruction rather than only measuring the ratio, because a compression figure means nothing without it. The number that carries the idea is 10 bits: the data was never 32 bits wide, it was 10 bits of information wearing a 32-bit costume, and the encoding stops storing a constant a hundred thousand times.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE INVENTORY · David Lee Wise (ROOT0), with AVAN