Abstract
Pointwise and uniform continuity become different properties by the exchange of a $\forall$ and an $\exists$. This paper takes a census of “continuous” in five senses and shows in numbers that the shape in which $\delta$ answers $\varepsilon$ is the classification itself. NO NEW MATHEMATICAL THEOREM IS CLAIMED — all of this is standard.
First, for $1/x$ on $(0,1)$ one has $\delta(x)=x^2\varepsilon/(1+x\varepsilon)$, the measurement agreeing with the closed form to every digit. At $x=0.002$ it is $4.0\times10^{-7}$ — the infimum falls to $0$ and the function is not uniformly continuous (Section 2).
Second, the fence is compactness. On $[a,1]$ the infimum is strictly positive whatever $a$ may be ($1.0\times10^{-7}$ even at $a=0.001$) (Section 3).
Third, and this is the core. The slope of $\log\delta$ against $\log\varepsilon$ is the reciprocal of the Hölder exponent — $1.0$ for $2x+1$, $1.00054$ for $x^2$, $\mathbf{2.0}$ for $\sqrt{x}$, $\mathbf{3.0}$ for $x^{1/3}$ (Section 4).
Fourth, some continuity is worse than every power. For $1/\log(1/x)$ the $\delta$ at $\varepsilon=0.05$ is $2.06\times10^{-9}$, agreeing to every digit with $e^{-1/\varepsilon}$ — uniformly continuous, and Hölder for no exponent at all (Section 5).
Fifth, and this is the brake. Uniform continuity is not absolute continuity. The Cantor function has slope $0$ across every gap and climbs by exactly $1$ (the gaps totalling $0.99999986$) (Section 6).
Scope (what is not claimed)
NO NEW MATHEMATICAL THEOREM IS CLAIMED — the $\varepsilon$-$\delta$ formulation, the Heine–Cantor theorem, Hölder continuity, and the failure of absolute continuity for the Cantor function are all standard. Nor is $\varepsilon$-$\delta$ criticised — what is shown is how much the definition is keeping apart. What is added is measuring the infimum of $\delta$ and exhibiting how it falls, confirming numerically that the slope of $\log\delta$ against $\log\varepsilon$ equals the reciprocal Hölder exponent, measuring a $\delta$ worse than every power, and setting the vanishing slope of the Cantor function beside the total length of its gaps.
1. Notation and position
A function is continuous at a point when for every $x$, for every $\varepsilon$, there is a $\delta$ such that $|y-x|<\delta$ forces $|f(y)-f(x)|<\varepsilon$. It is uniformly continuous when for every $\varepsilon$ there is a $\delta$ such that for every $x$ the same holds. The words are identical; only the order of $\forall x$ and $\exists\delta$ differs. Every $\delta$ here was measured by bisection.
Position: Paper 171 wrote that the fence around least squares is linearity, and Paper 173 that Kodaira's fence is integrality. This is the third of them — the fence is compactness. Paper 154 treated the order of limit operations, while this paper treats the order of quantifiers — different orders, and so no collision.
2. Exchanging $\forall$ and $\exists$ gives a different property
Verification 2.1: Take $f(x)=1/x$ on the open interval $(0,1)$ and measure $\delta$ at each point for $\varepsilon=0.1$.
| $x$ | measured $\delta(x)$ | $x^2\varepsilon/(1+x\varepsilon)$ |
|---|---|---|
| $0.9$ | $0.0743119266055$ | —— |
| $0.5$ | $0.0238095238095$ | $0.0238095238095$ |
| $0.2$ | $0.00392156862745$ | —— |
| $0.05$ | $0.000248756218905$ | $0.000248756218905$ |
| $0.01$ | $9.99000999001\times10^{-6}$ | $9.99000999001\times10^{-6}$ |
| $0.002$ | $\mathbf{3.99920015997\times10^{-7}}$ | —— |
Measurement and closed form agree to every digit. (QED)
Observation 2.2 As $x$ approaches $0$ the $\delta$ falls like $x^2$. No single $\delta$ serves every $x$ at once.
Observation 2.3 So $f$ is continuous at every point and not uniformly continuous. Nothing about the function changed — the quantifiers were reordered.
3. The fence is compactness
Verification 3.1: Put the same function on the closed interval $[a,1]$ and take the infimum of $\delta$ over $401$ points.
| $a$ | infimum of $\delta$ |
|---|---|
| $0.5$ | $0.0263157894737$ |
| $0.1$ | $0.0010101010101$ |
| $0.01$ | $1.001001001\times10^{-5}$ |
| $0.001$ | $\mathbf{1.00010001\times10^{-7}}$ |
Whatever $a$ may be, the infimum is strictly positive. (QED)
Observation 3.2 This is the Heine–Cantor theorem — on a compact set continuity is automatically uniform. The fence is compactness, and the open end at $0$ was the only thing breaking it.
4. The core: the shape in which $\delta$ answers $\varepsilon$ is the classification
Verification 4.1: For uniformly continuous functions, take the slope of $\log\delta$ against $\log\varepsilon$.
| function | $\delta$ at $\varepsilon=10^{-2},10^{-3},10^{-4}$ | slope | kind |
|---|---|---|---|
| $2x+1$ | $0.005,\ 0.0005,\ 5\times10^{-5}$ | $\mathbf{1.0}$ | Lipschitz |
| $x^2$ | $0.00501256,\ 0.000500125,\ 5.00013\times10^{-5}$ | $1.00054$ | Lipschitz |
| $\sqrt{x}$ | $10^{-4},\ 10^{-6},\ 10^{-8}$ | $\mathbf{2.0}$ | Hölder $1/2$ |
| $x^{1/3}$ | $10^{-6},\ 10^{-9},\ 10^{-12}$ | $\mathbf{3.0}$ | Hölder $1/3$ |
(QED)
Observation 4.2 The slope is the reciprocal of the Hölder exponent. One for Lipschitz, two for Hölder one half, three for one third.
Observation 4.3 How $\delta$ answers $\varepsilon$ is not a detail of a proof — it is the classification. To say “continuous” is already to say something about how fast the answer comes.
5. Worse than every power, and still uniform
Verification 5.1: Take $f(x)=1/\log(1/x)$, with $f(0)=0$, on $[0,1/2]$.
| $\varepsilon$ | $\delta$ near $0$ | $e^{-1/\varepsilon}$ |
|---|---|---|
| $0.5$ | $0.1353352832$ | $0.1353352832$ |
| $0.2$ | $0.006737946999$ | $0.006737946999$ |
| $0.1$ | $4.539992976\times10^{-5}$ | $4.539992976\times10^{-5}$ |
| $0.05$ | $\mathbf{2.061153622\times10^{-9}}$ | $\mathbf{2.061153622\times10^{-9}}$ |
They agree to every digit. (QED)
Observation 5.2 The $\delta$ falls like $e^{-1/\varepsilon}$ — faster than $\varepsilon^k$ for every $k$. Uniformly continuous, and Hölder for no exponent at all. “Uniform” does not mean “with a power-law modulus”.
6. The brake: uniform is not absolute
Observation 6.1 The Cantor function is continuous on $[0,1]$, hence uniformly so, and it climbs from $0$ to $1$.
Verification 6.2: The difference quotient across the removed middles.
| gap | $f$ at the left | $f$ at the right | slope |
|---|---|---|---|
| $1/3$ to $2/3$ | $0.5$ | $0.5$ | $\mathbf{0}$ |
| $1/9$ to $2/9$ | $0.25$ | $0.25$ | $\mathbf{0}$ |
| $7/9$ to $8/9$ | $0.75$ | $0.75$ | $\mathbf{0}$ |
| $1/27$ to $2/27$ | $0.125$ | $0.125$ | $\mathbf{0}$ |
Verification 6.3: The total length of those gaps.
| $\sum_{n\ge1}2^{n-1}/3^n$ to $40$ levels | $\mathbf{0.99999986434341}$ |
|---|---|
| $f(1)-f(0)$ | $\mathbf{1.0}$ |
(QED)
Observation 6.4 The gaps fill measure $1$. So $f'=0$ almost everywhere, and yet $f$ climbs by exactly $1$. The integral of $f'$ is $0$ and the rise is $1$ — the fundamental theorem fails.
Observation 6.5 Uniform continuity was not enough. What is needed is absolute continuity, and the Cantor function is the standard witness that the two differ.
Observation 6.6 This does not contradict Paper 168. What that paper said, that the gauge integral recovers every derivative, is about functions differentiable everywhere. The Cantor function is not differentiable on the Cantor set, so it fails that theorem's hypothesis. The fence holds here too.
Observation 6.7 And “almost everywhere” is to be read in the sense of Paper 176 — what vanishes here is measure, not category.
7. The census
Observation 7.1
| kind | quantifiers or shape | separator | witness |
|---|---|---|---|
| pointwise | $\forall x\,\forall\varepsilon\,\exists\delta$ | $\delta$ may depend on $x$ | $1/x$ on $(0,1)$ |
| uniform | $\forall\varepsilon\,\exists\delta\,\forall x$ | $\delta$ does not | automatic on a compact set |
| Hölder | $\delta\propto\varepsilon^{1/\alpha}$ | answers by a power | $\sqrt{x}$, slope $2$ |
| Lipschitz | $\delta\propto\varepsilon$ | answers linearly | slope $1$ |
| absolute | stated over sums of intervals | the fundamental theorem holds | the Cantor function |
Observation 7.2 These five are a staircase — but each step is there for a different reason. The first two are separated by the order of quantifiers, the second and third by how fast the answer comes, the fourth and fifth by whether the fundamental theorem holds.
8. Conclusion
- The infimum of $\delta$ falls to $0$ (Ver. 2.1): $4.0\times10^{-7}$ at $x=0.002$, matching the closed form to every digit.
- The fence is compactness (Ver. 3.1): on $[a,1]$ the infimum is positive whatever $a$ is.
- The slope is the reciprocal Hölder exponent (Ver. 4.1): $1.0$, $2.0$, $3.0$.
- Some uniform continuity beats every power (Ver. 5.1): $\delta\simeq e^{-1/\varepsilon}$.
- Uniform is not absolute (Ver. 6.2, 6.3): slope $0$ throughout, and a rise of $1$.
References
- Bolzano, B. (1817). Rein analytischer Beweis des Lehrsatzes, dass zwischen je zwey Werthen, die ein entgegengesetztes Resultat gewähren, wenigstens eine reelle Wurzel der Gleichung liege. Prague: Gottlieb Haase.
- Cauchy, A.-L. (1821). Cours d'analyse de l'École Royale Polytechnique. Paris: Imprimerie Royale.
- Heine, E. (1872). Die Elemente der Functionenlehre. Journal für die reine und angewandte Mathematik, 74.
- Cantor, G. (1883). Über unendliche, lineare Punktmannichfaltigkeiten V. Mathematische Annalen, 21.
- Vitali, G. (1905). Sulle funzioni integrali. Atti della Reale Accademia delle Scienze di Torino, 40.
- Lebesgue, H. (1904). Leçons sur l'intégration et la recherche des fonctions primitives. Paris: Gauthier-Villars.
- Yamagishi, Y. (2026). “Dimension” Is Not One Thing [Paper 94]. Zenodo. 10.5281/zenodo.21758581
- Yamagishi, Y. (2026). Modules Are Not Walls but Limits [Paper 113]. Zenodo. 10.5281/zenodo.21962996
- Yamagishi, Y. (2026). A Limit Without Its Order Is Not a Quantity [Paper 154]. Zenodo. 10.5281/zenodo.22084647
- Yamagishi, Y. (2026). There Are Spaces With No Points at All [Paper 155]. Zenodo. 10.5281/zenodo.22084848
- Yamagishi, Y. (2026). What Lebesgue Bought Was Not a Bigger Class but Closure under Limits [Paper 168]. Zenodo. 10.5281/zenodo.22093320
- Yamagishi, Y. (2026). The Fence around Least Squares Is Linearity, Not Normality [Paper 171]. Zenodo. 10.5281/zenodo.22094287
- Yamagishi, Y. (2026). Kodaira's Fence Is Integrality, Not Positivity [Paper 173]. Zenodo. 10.5281/zenodo.22094507
- Yamagishi, Y. (2026). “Symmetry Breaking” Is Not One Word [Paper 175]. Zenodo. 10.5281/zenodo.22097873
Reference
Yamagishi, Y. (2026). The Order of the Quantifiers Is the Whole Content [Paper 177]. YOI: https://www.yamagishiyuuki.com/yoi/177
DOI: 10.5281/zenodo.22098103
ORCID: 0009-0006-8587-3450
PDF (English) · PDF (Japanese) · Date proof of the English PDF (OpenTimestamps .ots) · SHA256SUMS
The record page for this paper (in Japanese, with the PDFs in each language and the date proof): YOI 177. This paper is a preprint and has not been peer reviewed.