The pricing schedule
The dynamics a growth law's limit comes out of, with the
ring taken out: five ingredients and a supply, the dial survey that leaves
one infinite coordinate standing, the covering rule that is the second
mechanism known to stop the openings, how big a finite image is,
and what the element world needs on top.
A growth law extends a modulus by the least move meeting a
structural demand (Growth), and what a run converges to
is a limit: one exponent per place — the primes of a ring past
Z, each carrying a degree, the degree of the irreducible for
polynomials and the logarithm of the norm for a number ring
(The image and limit). A move that seats a place
not yet seated is an opening; deepening one already seated is a
repeat move, and a repeat advances a counter, the tick
T, against which every price is read — an exponent at or under the
tick buys nothing new, so a repeat must climb past it. A run locks
when it stops opening new places for good, and the repeated move it makes
forever after is its recurrent move. In a limit, a coordinate carried
up without bound is a runaway; a coordinate left standing above
exponent 1 by a move it can never afford again is a strand.
Take the ring out
Those shapes were derived in rings — a function field, a quadratic
number ring, an elliptic curve's coordinate ring — and the derivation
needs none of them.
The limit belongs to
the schedule, not the ring rule
Take the ring out and four ingredients remain: items carrying an
integer DEGREE; one GLOBAL CLOCK multiplying the tick by b
whenever any item is deepened; a PRICE f(degree, staleness),
staleness
being how far the tick has moved since that item last did; and a
fresh-opening DISCOUNT spendable m times at each degree, the
least degree born with its discount already spent — a degree whose
discount is gone being covered, so a fresh opening goes to the
least uncovered degree. The ring enters
only as a SUPPLY — how many items each degree holds. Dial each as
far as the move model admits — b at 3/2, 2, 3 and 4, the
degree's exponent at 0, 1 and 2 and an additive form, one to three
discounts, every born-spent set — and ONE INFINITE COORDINATE
survives every dial but one. The dial that kills it is the
degree-blind price, which cannot tell a cheap item from a dear one —
and only where the clock's steps are wider than 1. At step 1 an
opening and a repeat cost the same, the tie deepens rather than
opens, and one item runs away regardless.
A FLAT SUPPORT — every coordinate but the deep one at exponent
1 — is the weaker half and does not survive as far, so
the two are separate claims. At b ≥ 3 the ceiling on the deep
item's degree — it is bounded by twice the least degree — has already
fallen to
the least degree while the clock still seats an item above it, so
that item is seated, clocked, then undercut and STRANDED above
exponent 1 forever — exactly one such item on every branch that
hands the clock on, and none on the branch whose first repeat was
also its last. The doubling clock of
the ideal limit is precisely
where that cannot happen.
Scope. Proved for the abstract schedule, and
a rule in range: 60 walks over ten schedules and six supplies, with
the abstract walker certified equal to the exact one in cost, in
every move type's multiplicity and in the covered set at 120 states
over each of six supplies. The dial survey is crossed with the
clock's landing rule rather than read beside it — 7 landing rules ×
8 dials × 2 clocks — and the single infinite coordinate holds at all
49 crossed cells whose price can see a degree UNDER THE GLOBAL
CLOCK. Give each item its own tick instead and the count standing
above exponent 1 moves with the dial and with the clock's step width
together, which is the mixed run of
the clock's own dial. Strands
arise on both sides by different mechanisms — there the ceiling falls under a seated
item, in a mixed run the item's own step widens under it.
verifiers:
explore_price_schedule.py,
explore_tick_pump.py
That dynamics has a name outside this subject. Size-based priority
with aging is a scheduling discipline in production use — its UNIX
form halves a usage counter every second, which is this clock at
b = 2 — and it has a deterministic limit analysis of its own:
Epema's decay-usage steady state (ACM TOCS 16(4), 1998)
partitions a fixed job set into monopolisers, a shared middle, and
classes that starve at share exactly zero. The claims here are
narrower than that overlap suggests: a size-discriminated ceiling in
closed form, under ADMISSION from a supply that never runs out, so the
support itself grows — which a fixed job set does not do. And on the
one question the two analyses share — how many jobs end up
monopolising — they part company. There, the boundaries of the
partition are not known beforehand, so the shares seem to admit no
closed form — the analysis says so itself, computes both by an
algorithm, and closes with the shares depending on the numbers
of jobs in the classes in an intricate way. Here the same
count has a law with no arithmetic in its statement — exactly one
runaway where the recurrent price is flat, one per block where it
climbs — at the block clock,
where the two cases and their mechanism live.
The other dial that stops the openings
A ring hands the schedule its clock one place at a time
(The clock), and what a place gives the dynamics
is its gap: the distance between the exponents the tick may stand at,
which is the price of a repeat move divided by the degree. A bounded gap
stops the openings and an unbounded one lets them run forever, which is the
one stop the four ingredients above can offer. Underneath the gap
sit two things the element world below needs — λ, the exponent of the
unit group, which is what a move has to move, and a place's door, the
least climb there that raises λ
against the whole state at once.
The discount carries a second dial,
hidden by the only setting ever run: “spendable m times at each
degree” presumes that an opening covers its OWN degree and nothing
else. Which degrees an opening covers is a covering rule, and it is a
fifth ingredient rather than a phrasing of the fourth.
A covering rule stops the
openings only by outrunning them rule
Dial the covering rule and one thing decides whether the openings stop
on their own: whether it can put degrees out of reach faster than the walk
opens them. A rule keyed to the SEATED SET never can: what it covers reads
only the degrees already opened, which no declined move changes, so the
cheapest fresh opening holds a constant price that the doubling clock
eventually undercuts. An opening covering its DIVISORS opens exactly the
degrees the bare rule opens, the least uncovered degree only ever
increasing; one covering its MULTIPLES opens the primes alone, and thins
the climb into going HIGHER in the same walk, the clock firing more often
for want of cheap openings. Such a rule is FED BY the openings it would
have to get ahead of. A rule keyed to the CLOCK outruns them outright, the
clock doubling where the openings step by one: at “covered while
d ≤ c·T” the openings are dead at c = 1
on every branch whose deep item has degree at most 2 — a home at
degree 3 climbs on at c = 1 and stops at c = 3/2 —
and at c = 1/2 dead only on the branch whose deep item has degree
1 — the degree-2 branch pays 2·(T/2) = T for its own repeat
and can still afford the least uncovered degree, so the threshold is the
deep item's degree against c. And a stop is read as a PRICE and
never as a silence: at the last state of each stopped walk the supply was
still offering a fresh opening — at degree 1025 or 2049 — and the walk
took a repeat costing 1024 instead.
Scope. A rule in range: 10 walks over 8 covering
rules against a supply carrying items at every degree to 4096, each state
asserted to sit far below that ceiling and to have a fresh opening left to
DECLINE, so that the supply is never what stops anything. The two
thresholds are exact by derivation and measured to a factor of two. The
seated-set half holds for the whole kind, the frozen frontier by argument
and the most generous member run: an opening covering every degree below
2d puts the covered set past every bound in four
openings — degrees 2, 4, 16, 65536 — and the walk still takes each next
opening the moment the doubling repeat price passes its frozen price.
Outrunning the DEGREES is not stopping the walk.
verifier:
explore_ladder_stop.py,
explore_seated_cover.py
One covering rule is not a dial at all, because a ring brings its own —
and the world that sprawls turns out to bring a rule that could not have
stopped it.
A function field's own
covering rule frees one degree rule
Read a function field's admission test as a covering rule and it is
DIVISIBILITY: based − 1 must divide the seated set's
invariant, and an invariant is an exponent there, so what accumulates
across the seated set is an LCM. At that reading nothing is ever covered
for free, at base 2 or base 3. At the strictly more generous PRODUCT
reading exactly one degree is freed — degree 6 at base 2, and nothing at
base 3 — and it is the classical exception that frees it:
2d − 1 carries a prime factor dividing no smaller member
at every d > 1 except d = 6 (Bang, 1886), so an LCM of
smaller terms can never supply what a fresh degree demands, while at
d = 6 the demand is 26 − 1 = 32·7, whose
factors a PRODUCT of smaller terms can supply and an LCM cannot. So a
function field's sprawl is a fact about its covering rule and not only
about its supply. The boundary is the load-bearing half: every reading
run here has integer degrees and one base, which is the function field's
shape,
while a number ring's degree is a logarithm of a norm and its test runs
over residue-field orders sharing no base, so nothing there plays Bang's
part. That exonerates the covering rule where the openings are known not
to stop; where they are known to stop, the exoneration comes from the
clock instead, whose gaps never grow, and has nothing to do with
covering.
Scope. A rule in range at four (reading, base)
rows, run on the walks above, beside a control reproducing from the
schedule side that the openable degrees are every degree. Bang's theorem
is an external input and is not re-derived here.
verifier:
explore_ladder_stop.py
How big a finite image is
Stopping the openings is not the only way a run ends with finitely many
futures — a supply with a TOP DEGREE simply runs OUT. How many futures is a
different question from whether they are finitely many, and these two
mechanisms answer it differently, which is most of what there is to say about
them. The set of limits a cheapest-move policy can reach is its image
(The image and limit), a set at all
only because moves can tie.
Both counts are set at the state a run starts from, the empty one. The
cheapest first move there is a comparison: the least degree the supply
offers fresh, at its discounted price, against the least degree born
covered, which has no discount left and pays the next price up. A degree
that WINS that comparison is a home — a
degree the deep item can settle at — and there are at most two, one fresh
and one covered, since the price rises with the degree at either entry
and
two of a kind can never tie. Whether there are two is a tie, and at the
linear price the tie is exact arithmetic: the next price up is twice the
discounted one, so the two sides draw exactly when the least fresh degree is
twice the least covered one. How many items the supply holds at a degree is
that degree's width.
What a finite image
is rule
Let the supply run out. The image is a SUM
over the homes of a PRODUCT — the home's own width times the width of every
other degree the walk can still open. For n items at every degree 1
to D that is nD−1·(n+1), which at
n = 2 and D = 8 reads 384 limits: the “+1” keeping
it off a power of 2 is a SECOND HOME and nothing else. Break the tie between
the two homes — gap the supply so the least fresh degree is no longer twice
the least covered one, steepen the price, or lift the supply's least degree
so the fresh side wins outright — and the sum has one term, a pure power of
the width: under a gapped supply 128 = 27 at width 2 and 243 =
35 at width 3. The product runs over the degrees the walk COULD
open rather than the ones present: park a degree above the home where the
walk steps past it and it stands empty at every reachable state, 16 against
the 32 a count of present degrees gives. Where the widths differ by degree
the product shows itself as one, 48 = 2·3·2·4.
Now stop the ladder
instead. The image is the PRODUCT over the homes alone of (width + 1), less
one for the state with no deep item yet, and the widths ABOVE the homes
never enter: two supplies agreeing at widths 2
and 3 on the homes and differing at every degree above both read
11 = 3·4 − 1. So a stop does not merely make an image finite, it makes it
SMALL — at most (n+1)2 − 1 for n the larger
home's width, whatever the supply, where an exhausted ladder multiplies in
a width for every degree it could ever reach. And a stopped image stands
above a bare count of its homes' widths for two reasons — the second home,
and the one degree a stop leaves OPTIONALLY open — which a tie-break
removes together, they being the same degree. At the linear price — the
degree's exponent 1 — a walk stops at its first clock or never, so every
opening a stopped walk makes happens at the first tick, and a degree opened
there is exactly a degree that won the comparison at the empty state, which
is what a home is.
The power-of-2 reading does not travel between the two mechanisms.
Under a stop the tied case reads 8 = 23 at width 2 and the
untied one 2, both powers; what separates them there is width 3,
15 = 3·5 against 3.
Scope. The exhausted law is DERIVED, from the homes
being the winners of that comparison and an exhausted state having one
shape per home; that every shape it allows is actually REACHED is measured
and not derived. A rule in range over nine families — seven untied across
three dials of the supply and the price, beside two tied controls — with
the decomposition asserted at every reachable state rather than read off
the count, and the covering-rule rig's own tied rows reproduced through
this rig's enumerator as a positive control, read first. The stopped law is
PROVED at the linear price, where the stop-at-the-first-clock step is
derived rather than measured, and a rule in range over eleven families with
none off the closed form — at ONE discount per degree and a doubling clock,
both load-bearing: more discounts break the step that leaves one flat item,
and a deep move's undercut threshold is (b−1)/b of the
tick, so the
threshold is the clock's as much as the degree's. And it is a closed form
only where the covering rule stops EVERY branch: the covered home's
threshold is half the fresh home's, so a rule set between them stops the
covered branch and leaves the image infinite on the fresh one, whose factor
is then not a count. Nothing here is a claim about a RING's finite image,
which the first mechanism cannot reach at all: a degree there is the
logarithm of a norm, so the supply never runs out and
the ladder stops on a flat recurrent price instead.
verifiers:
explore_void_untie.py,
explore_stopped_untie.py
The element world is a schedule too
Growing by elements — every state and every move generated by a single
element — adds one ingredient, and it is the only one the
dials above do not reach. A move deepens a core — one place
raised to a power — and what a single element generates must have
trivial ideal class, the class group being the finite group that
measures how far a ring's ideals are from being generated by one element,
so the move carries a second factor: the
rider, the cheapest product of places whose class cancels the
core's. The rider's COST is a lookup on the core's own degree and class
and on that core's door. Its EFFECT is not
a price at all — the move raises exponents at
places the core is not, and a family whose move raises one item's
exponent has no member that raises several, at any price. So the cost
rejoins the schedule and the effect stays outside it.
The element schedule,
in both characteristics rule
Strip the ring again, and what the ideal world read as a supply by
degree the element world reads as a supply matrix — how many
places carry each (degree, class) pair, over the ideal classes as a
finite group with its addition table. Hand a walker that, plus the
clock's reading at each state. Recompute the rider and its cost from the
matrix alone, as a shortest path over that group with each class
weighted by its cheapest place, and the walker's menu — the
cheapest admissible moves at a state, one entry where the greedy move
is unique and several where it ties — is the ring's own, entry for
entry, at every state. It holds over six function-field rings, where a
place's degree is an integer, and over two quadratic number rings,
where a place's degree is the logarithm of its norm, so that costs
which added as degrees now multiply as norms. The least cost stops being a
least total degree and becomes a least norm PRODUCT, its weights no
longer integers, and nothing breaks: the two arguments that use it —
the shortest path, and the bound saying a core bought past its door is
never cheaper — need only that it is a minimum over an ordered monoid
with a monotone operation, and the logarithm carries one such monoid
to the other.
The matrix does not give everything. A function field's places
share one clock, the doubling tick; a number ring gives each place its
own ladder — the set of depths at which that place's own
λ moves, its tail gap fixed by that place's ramification,
how many times the rational prime beneath it repeats there
(The clock) — so no single
tick supplies the doors, and the matrix, being a COUNT of how many
places carry each pair, holds no per-place ladder to read one from. The ring hands a
LADDER COLUMN over beside it, and that is the whole price of the
crossing. Norm and gap together come close:
for a place of norm N at depth a with gap g,
rad(N) being the rational prime beneath it. That is the ring's
own λ at every place except the ones whose ladder carries a
leading stretch of wider gaps
(The clock), where the
principal units are still squaring rather than stepping — over these two
rings the places over 2. So the ring enters as a
supply matrix and a ladder column — one entry per place, and over a
function field the same entry throughout — and the ladder is what
hands the walker its doors wherever the state does not widen one,
which over these two rings is everywhere.
The bundle's form also says why an element lock's recurrent price
is FLAT,
which the ideal world's own argument could not reach: that one is
about a single place's valuation, and an element move seats several
places at once. It does not have to reach several. The recurrent move is the core at a fixed power
together with its rider, so the price FACTORS — the first factor flat
because deepening a place carries the state's valuation to just below
the new depth, leaving the next door at the constant gap, and the
second a lookup on a finite group, fixed once the door is. The element world's flat price is the
ideal mechanism composed with the rider being a price.
Scope. The cost monoid's carry-over is a
derivation; the menu equality is a rule in range on both sides, the
walker handed the clock's reading at each state — the tick over a
function field, the door itself over a number ring — so what is under
test is the bundle and the cost, never the ladder. Six function-field
rings, menus compared entry for entry over the whole branch tree out
to four moves and nine moves of the greedy line each, none differing;
and two quadratic number rings, 675 states — every element seed of
norm ≤ 40 walked twelve moves, plus the whole branch tree out to four
moves — again none differing, and of the 344 and 347 menu entries, 103
and 307 seat a bundle rather than a single place. The ladder column is
read at the 35 places of norm ≤ 60 over depths 1..14, as a
biconditional against the criterion for a leading stretch
(The clock), so that an
agreement at one of the places carrying one would count against it too: 32
agree, 3 part from depth 3 on, 35 of 35 on the criterion. Each class's cheapest rider is unique
at both number rings — proved by an enumeration clearing the
Minkowski bound — and at every
function-field ring run; where a class does have several, the width
decides: offering them all keeps the orbit count, and taking a
canonical one keeps it at width 2 and breaks it at width 3. The recurrent tails are read at 53 seeds with 40 moves past
each lock, 0 non-flat.
verifiers:
explore_class_schedule.py,
explore_element_schedule_nf.py
The rider rejoins the schedule as a cost, then, and the effect it has
outside one is not only that it seats several places at once. It also
reaches the prices.
A rider discounts a door,
and never the clock rule
A rider is not priced only at its own degree. It raises the exponent at
every place it seats, and a place's door falls by exactly what its
exponent gains, with no repeat at all — so a FRESH element move can
drop the menu's minimum, which is an effect the ideal world has no move
for. A tie is therefore breakable from outside itself: the
cheapest move at a later state need be neither member of it, and the loser
is priced out rather than beaten on its own terms. Every fall in a menu
minimum along these walks sits at a place the move that caused it seats —
its core, or its rider — and at none anywhere else. Over six
function-field rings the element walks read 18 falls in 1,370 moves, 14 at
a core and 4 at a rider; the ideal walks read 37 in 1,632, all at a core,
there being no riders there to read.
Scope. A rule in range: every fall in the menu
minimum attributed along ideal and element walks over six rings, 0 outside
the causing move's own support. Two families of declined moves that a
comparison of door prices leaves uncovered — the reading behind
the separation lemma — sit at
a fallen minimum at 74 of 74 and 32 of 32, which is what a discounted door
accounts for and a comparison of prices at one state cannot. That a rider
never advances the tick is read across 2,351 moves and 44,828 place
readings, 0 moving it.
verifier:
explore_undercut.py
What the clock hands over
The clock the prices above are read against is a
ring's, and it is a local invariant: whether the openings ever stop is the
characteristic — equal against mixed — and within the side that stops,
ramification sets where. A place with a leading stretch of wider gaps
hands the dynamics two numbers rather than one, the sup gap deciding where
the walk stops and the tail gap what the runaway pays forever. Both are
The clock. And a ring
sits at neither end of the family's clock dial but at a cell between them,
one tick per rational prime with a ladder per place, where a single deep
coordinate is the flat-cost corner of a law that otherwise reads one runaway
per rational prime — The clock dial.
What that dichotomy decides at the level of the machine — the same
apparatus read as a computation, its depth face one borrow short of
Turing-complete and its element face universal bare — is the
Computation section.