Little Lipschitz Space Overview

• Little Lipschitz space is a refinement of Lipschitz spaces, defined via functions whose local Lipschitz constants vanish at small scales, capturing a notion of local flatness.
• It extends classical theory to various settings such as ultrametric spaces, rooted trees, and locally finite graphs by adapting the vanishing condition to the underlying geometry.
• The space plays a crucial role in establishing duality with Lipschitz-free spaces and in exploring bidualities and predual structures, enhancing metric and Banach space analysis.

The little Lipschitz space is a Banach-space-theoretic and metric-analytic refinement of the Lipschitz space, designed to isolate functions whose Lipschitz behavior vanishes in an appropriate asymptotic regime. On a pointed metric space $(M,d,0)$, the classical form consists of functions in $\operatorname{Lip}_0(M)$ whose local Lipschitz constants tend to $0$ at small scales; in other frameworks the vanishing condition is adapted to the geometry, for example by requiring flatness at infinity on noncompact ultrametric spaces or vanishing edge increments on rooted trees and locally finite graphs (Jiménez-Vargas et al., 2014, Abrahamsen et al., 26 Sep 2025, Colonna et al., 2024, Issa-Barbará et al., 14 Feb 2026). This suggests that “little Lipschitz space” is best understood as a family of closely related constructions rather than a single uniform definition.

1. Classical definition and principal variants

For a pointed compact metric space $(X,d,0)$, the Banach space

$$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$

is endowed with the Lipschitz norm

$$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$

Its little Lipschitz subspace is

$$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$

so the local Lipschitz constants vanish at small scales (Jiménez-Vargas et al., 2014). On compact spaces, the same condition is described as local flatness, equivalently

$$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$

and this matches the usual little Lipschitz condition $\omega_f(t)/t\to 0$, where $\omega_f(t)$ is the modulus of continuity (Aliaga, 1 Jun 2026).

For general pointed metric spaces, the ultrametric framework of “Duality of Lipschitz-free spaces over ultrametric spaces” replaces the compact small-scale formulation by a two-part requirement. For $\operatorname{Lip}_0(M)$0, define

$\operatorname{Lip}_0(M)$1

The paper uses the conditions of being uniformly locally flat and flat at infinity, and defines

$\operatorname{Lip}_0(M)$2

If $\operatorname{Lip}_0(M)$3 is proper, then $\operatorname{Lip}_0(M)$4 coincides with the standard $\operatorname{Lip}_0(M)$5 (Abrahamsen et al., 26 Sep 2025).

Several specialized models retain the same vanishing-Lipschitz philosophy while changing the ambient geometry. On rooted trees, the little Lipschitz space $\operatorname{Lip}_0(M)$6 consists of functions $\operatorname{Lip}_0(M)$7 such that

$\operatorname{Lip}_0(M)$8

where $\operatorname{Lip}_0(M)$9 is the parent of $0$0 (Colonna et al., 2024). On an infinite connected locally-finite graph $0$1, the little Lipschitz space is

$0$2

and this definition is independent of the choice of basepoint (Issa-Barbará et al., 14 Feb 2026). For vector-valued mappings on a compact metric space,

$0$3

so the local Lipschitz constant tends to $0$4 at small scales in the norm of $0$5 (Izumi, 2020).

A distinct but related terminology appears in the pointwise theory of little Lipschitz mappings. There,

$0$6

or, in a later metric-derivative formulation,

$0$7

These are pointwise quantities rather than Banach subspaces, but they formalize the same small-scale vanishing or local finiteness principle (Malý et al., 2018, Maslyuchenko et al., 25 Sep 2025).

2. Duality, Lipschitz-free spaces, and biduality

The little Lipschitz space is tightly linked to the linearization of metric spaces through Lipschitz-free spaces. For a pointed metric space $0$8, the Lipschitz-free space $0$9 is the closed linear span of the evaluation map $(X,d,0)$0, and one has the canonical duality

$(X,d,0)$1

(Aliaga, 1 Jun 2026). In the compact pointed setting, Johnson’s predual $(X,d,0)$2 yields the canonical isometric isomorphism $(X,d,0)$3 (Jiménez-Vargas et al., 2014).

A central question is when $(X,d,0)$4 is isometrically isomorphic to $(X,d,0)$5. For pointed compact metric spaces, this occurs exactly when the closed unit ball $(X,d,0)$6 is dense in $(X,d,0)$7 with respect to the topology of pointwise convergence $(X,d,0)$8, equivalently with respect to the weak* topology on bounded sets (Jiménez-Vargas et al., 2014). The same paper proves that for every $(X,d,0)$9,

$$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$0

isometrically, giving the classical snowflake biduality result (Jiménez-Vargas et al., 2014).

For compact purely $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$1-unrectifiable spaces, the little Lipschitz space has an even stronger role. The equivalences reviewed in “Lipschitz-free spaces and purely 1-unrectifiable metric spaces” assert that, for compact $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$2, the following are equivalent: $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$3 is purely $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$4-unrectifiable; $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$5 separates points uniformly; $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$6; $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$7 is a dual Banach space; $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$8 has the Radon–Nikodým property; $$\mathrm{Lip}_0(X,d):=\{f:X\to K:\ f(0)=0,\ \mathrm{Lip}_d(f)<\infty\}$$9 has the Schur property; and $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$0 does not contain an isomorphic copy of $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$1 (Aliaga, 1 Jun 2026). In this compact regime, locally flat functions are weak* dense in $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$2, and every $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$3 attains its norm on a molecule $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$4 (Aliaga, 1 Jun 2026).

The vector-valued theory extends this bidual perspective. If $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$5 is compact, $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$6 separates points uniformly, and either $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$7 or $$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$8 has the approximation property, then

$$\|f\|_{\mathrm{Lip}}=\sup_{x\neq y}\frac{|f(x)-f(y)|}{d(x,y)}.$$9

isometrically (Izumi, 2020). The canonical isometric isomorphism is given pointwise by

$$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$0

where $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$1 is the evaluation functional at $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$2 (Izumi, 2020). This identifies the bidual of the vector-valued little Lipschitz space with all $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$3-valued Lipschitz functions rather than merely with the image of $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$4, showing that little Lipschitz regularity is generally not stable under bidualization (Izumi, 2020).

3. Ultrametric spaces, $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$5-ideals, and canonical preduals

Ultrametric spaces furnish a particularly rigid setting in which the little Lipschitz space acquires strong geometric structure. For a complete separable ultrametric space $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$6, the main theorem of “Duality of Lipschitz-free spaces over ultrametric spaces” states that the following are equivalent: $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$7 is a dual Banach space; $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$8 is $$f\in \mathrm{lip}_0(X,d)\quad\Longleftrightarrow\quad \lim_{\delta\to 0^+}\ \sup_{0<d(x,y)\le \delta}\frac{|f(x)-f(y)|}{d(x,y)}=0,$$9-complemented in $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$0; and $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$1 is spherically complete (Abrahamsen et al., 26 Sep 2025). In particular, proper ultrametric spaces are spherically complete, hence $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$2 is a dual space (Abrahamsen et al., 26 Sep 2025).

Within this setting, the relevant little Lipschitz space is

$$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$3

The decisive result is that for every ultrametric space $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$4, $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$5 is an $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$6-ideal in $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$7 (Abrahamsen et al., 26 Sep 2025). The proof uses the $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$8-ball property and an ultrametric partition scheme at scales $$\lim_{r\downarrow 0}\ \sup_{x\in M}\operatorname{Lip}\big(f|_{B(x,r)}\big)=0,$$9, constructing piecewise-constant approximants $\omega_f(t)/t\to 0$0 and exploiting the strong triangle inequality

$\omega_f(t)/t\to 0$1

to obtain the required uniform Lipschitz estimates (Abrahamsen et al., 26 Sep 2025).

The $\omega_f(t)/t\to 0$2-ideal statement interacts with a canonical predual $\omega_f(t)/t\to 0$3, defined from a dense sequence $\omega_f(t)/t\to 0$4 and $\omega_f(t)/t\to 0$5-Lipschitz retractions $\omega_f(t)/t\to 0$6. When $\omega_f(t)/t\to 0$7 is separable and spherically complete, every $\omega_f(t)/t\to 0$8 attains its norm and $\omega_f(t)/t\to 0$9 is norming, hence $\omega_f(t)$0; in the proper ultrametric case, $\omega_f(t)$1 (Abrahamsen et al., 26 Sep 2025). The same paper proves that for separable spherically complete ultrametric $\omega_f(t)$2, this predual $\omega_f(t)$3 is $\omega_f(t)$4-embedded if and only if $\omega_f(t)$5 is proper (Abrahamsen et al., 26 Sep 2025).

These facts give a particularly sharp picture in standard ultrametric examples. The $\omega_f(t)$6-adic integers $\omega_f(t)$7, the $\omega_f(t)$8-adic numbers $\omega_f(t)$9, and the Cantor set with the standard ultrametric are proper, so $\operatorname{Lip}_0(M)$00 is a dual space, $\operatorname{Lip}_0(M)$01 is an $\operatorname{Lip}_0(M)$02-ideal in $\operatorname{Lip}_0(M)$03, and in the proper case it is the $\operatorname{Lip}_0(M)$04-embedded predual of $\operatorname{Lip}_0(M)$05 (Abrahamsen et al., 26 Sep 2025). A plausible implication is that ultrametric geometry is exceptionally compatible with the Banach-space geometry of free spaces because spherical completeness, properness, and the $\operatorname{Lip}_0(M)$06-ideal structure reinforce one another.

The ultrametric theory also settles a negative problem outside this class. The paper answers a question posed by Werner by exhibiting a compact non-ultrametric space $\operatorname{Lip}_0(M)$07 for which $\operatorname{Lip}_0(M)$08 is not an $\operatorname{Lip}_0(M)$09-ideal in $\operatorname{Lip}_0(M)$10 (Abrahamsen et al., 26 Sep 2025). This shows that the $\operatorname{Lip}_0(M)$11-ideal property is a genuinely ultrametric phenomenon rather than a universal feature of little Lipschitz spaces.

4. Local flatness, unrectifiability, and metric structure

On compact metric spaces, the little Lipschitz space is closely tied to the absence of rectifiable one-dimensional structure. A metric space $\operatorname{Lip}_0(M)$12 is purely $\operatorname{Lip}_0(M)$13-unrectifiable if for every subset $\operatorname{Lip}_0(M)$14 and every Lipschitz map $\operatorname{Lip}_0(M)$15, one has $\operatorname{Lip}_0(M)$16; equivalently, $\operatorname{Lip}_0(M)$17 contains no bi-Lipschitz copy of a compact subset $\operatorname{Lip}_0(M)$18 with $\operatorname{Lip}_0(M)$19 (Aliaga, 1 Jun 2026). In this setting, local flatness is not merely a regularity condition but the mechanism that drives weak* density, predual identification, and linear properties of $\operatorname{Lip}_0(M)$20 (Aliaga, 1 Jun 2026).

The compact equivalence theorem reviewed in the expository manuscript states that for compact $\operatorname{Lip}_0(M)$21, pure $\operatorname{Lip}_0(M)$22-unrectifiability is equivalent to uniform point separation by $\operatorname{Lip}_0(M)$23: given $\operatorname{Lip}_0(M)$24 and $\operatorname{Lip}_0(M)$25, there exists $\operatorname{Lip}_0(M)$26 with $\operatorname{Lip}_0(M)$27 and

$\operatorname{Lip}_0(M)$28

It is also equivalent to $\operatorname{Lip}_0(M)$29, to $\operatorname{Lip}_0(M)$30 being a dual Banach space, and to the Radon–Nikodým and Schur properties of $\operatorname{Lip}_0(M)$31 (Aliaga, 1 Jun 2026). Conversely, if $\operatorname{Lip}_0(M)$32 contains a curve fragment, then locally flat functions cannot uniformly separate points along that fragment, so $\operatorname{Lip}_0(M)$33 is not weak* dense in $\operatorname{Lip}_0(M)$34 and $\operatorname{Lip}_0(M)$35 cannot serve as a predual of $\operatorname{Lip}_0(M)$36 (Aliaga, 1 Jun 2026).

Classical examples illustrate both sides of this dichotomy. Compact null $\operatorname{Lip}_0(M)$37 sets in $\operatorname{Lip}_0(M)$38, including Cantor-type sets, satisfy uniform separation by $\operatorname{Lip}_0(M)$39, and Godard’s theorem gives $\operatorname{Lip}_0(M)$40 (Aliaga, 1 Jun 2026). By contrast, for $\operatorname{Lip}_0(M)$41 with the Euclidean metric, $\operatorname{Lip}_0(M)$42, hence $\operatorname{Lip}_0(M)$43, while $\operatorname{Lip}_0(M)$44, so Schur and the Radon–Nikodým property fail (Aliaga, 1 Jun 2026).

The snowflake construction provides a large positive class. For $\operatorname{Lip}_0(M)$45 with $\operatorname{Lip}_0(M)$46, every $\operatorname{Lip}_0(M)$47 is uniformly locally flat for $\operatorname{Lip}_0(M)$48, and the density criterion yields the isometric identification of $\operatorname{Lip}_0(M)$49 with $\operatorname{Lip}_0(M)$50 (Aliaga, 1 Jun 2026, Jiménez-Vargas et al., 2014). This suggests that snowflaking destroys rectifiable structure in exactly the way needed for the little Lipschitz space to become large enough to recover the full dual geometry.

5. Discrete models on rooted trees and locally finite graphs

In discrete geometry, the little Lipschitz space is reformulated in terms of edge differences. On a rooted, locally finite tree $\operatorname{Lip}_0(M)$51 with root $\operatorname{Lip}_0(M)$52, the Lipschitz space $\operatorname{Lip}_0(M)$53 consists of functions with bounded discrete derivative

$\operatorname{Lip}_0(M)$54

The little Lipschitz space is

$\operatorname{Lip}_0(M)$55

it is separable, and the finitely supported functions are dense in it (Rivera-Guasco et al., 2019). The dual identifications

$\operatorname{Lip}_0(M)$56

hold with pairing

$\operatorname{Lip}_0(M)$57

(Rivera-Guasco et al., 2019).

The same space appears in later work under equivalent norms. On a rooted tree,

$\operatorname{Lip}_0(M)$58

with norm

$\operatorname{Lip}_0(M)$59

and $\operatorname{Lip}_0(M)$60 is isometrically isomorphic to $\operatorname{Lip}_0(M)$61 via the difference map

$\operatorname{Lip}_0(M)$62

For the path graph $\operatorname{Lip}_0(M)$63, this becomes

$\operatorname{Lip}_0(M)$64

(López-Martínez, 5 May 2025). This provides an explicit linear model of little Lipschitz spaces as $\operatorname{Lip}_0(M)$65-type sequence spaces.

Operator theory on these spaces is unusually concrete. The forward shift is bounded on $\operatorname{Lip}_0(M)$66 and on $\operatorname{Lip}_0(M)$67, and on a leafless tree it is an isometry; the backward shift is bounded exactly when the tree is homogeneous by sectors, and on a homogeneous tree of order $\operatorname{Lip}_0(M)$68,

$\operatorname{Lip}_0(M)$69

On $\operatorname{Lip}_0(M)$70, the backward shift is hypercyclic precisely when it is bounded and the tree has no free ends (Rivera-Guasco et al., 2019). For composition operators $\operatorname{Lip}_0(M)$71 on $\operatorname{Lip}_0(M)$72, boundedness is determined by Lipschitz properties of $\operatorname{Lip}_0(M)$73, and on the rooted path $\operatorname{Lip}_0(M)$74 hypercyclicity and the Frequent Hypercyclicity Criterion are equivalent to the growth condition

$\operatorname{Lip}_0(M)$75

(Colonna et al., 2024, López-Martínez, 5 May 2025).

A related discrete construction is the little Lipschitz space $\operatorname{Lip}_0(M)$76 on an infinite connected locally-finite graph $\operatorname{Lip}_0(M)$77. Here

$\operatorname{Lip}_0(M)$78

and

$\operatorname{Lip}_0(M)$79

The finitely supported functions are dense in $\operatorname{Lip}_0(M)$80, so $\operatorname{Lip}_0(M)$81 is separable, whereas $\operatorname{Lip}_0(M)$82 is not separable (Issa-Barbará et al., 14 Feb 2026). Multiplication operators $\operatorname{Lip}_0(M)$83 are bounded on $\operatorname{Lip}_0(M)$84 and on $\operatorname{Lip}_0(M)$85 exactly when $\operatorname{Lip}_0(M)$86 and

$\operatorname{Lip}_0(M)$87

and their spectrum is

$\operatorname{Lip}_0(M)$88

(Issa-Barbará et al., 14 Feb 2026).

These discrete models clarify a potential misconception. On trees and graphs, the adjective “little” does not refer to the limit $\operatorname{Lip}_0(M)$89 of a continuous metric modulus; instead it is encoded by vanishing edge oscillation at infinity (Colonna et al., 2024, Issa-Barbará et al., 14 Feb 2026). This suggests that the correct invariant notion is vanishing local oscillation relative to the geometry at hand.

6. Pointwise little Lipschitz derivatives, extensions, and mapping theory

Beyond Banach subspaces, little Lipschitz regularity is also studied pointwise. In “Mapping Analytic sets onto cubes by little Lipschitz functions”, a mapping $\operatorname{Lip}_0(M)$90 is called little Lipschitz if

$\operatorname{Lip}_0(M)$91

and every little Lipschitz mapping is continuous (Malý et al., 2018). The paper proves that if an analytic metric space $\operatorname{Lip}_0(M)$92 satisfies $\operatorname{Lip}_0(M)$93, then there exists a little Lipschitz surjective mapping

$\operatorname{Lip}_0(M)$94

(Malý et al., 2018). The proof proceeds through large compact subsets that are Lipschitz-equivalent to ultrametric spaces, monotone-space Hölder surjections onto $\operatorname{Lip}_0(M)$95, a $\operatorname{Lip}_0(M)$96-Hölder Peano curve, and an extension theorem for little Lipschitz maps on closed subsets (Malý et al., 2018).

The extension theorem is structurally significant. If $\operatorname{Lip}_0(M)$97 is closed, every little Lipschitz function $\operatorname{Lip}_0(M)$98 admits an extension $\operatorname{Lip}_0(M)$99 such that $0$00, $0$01 is little Lipschitz on $0$02, and $0$03 is locally Lipschitz on $0$04, hence little Lipschitz on all of $0$05 (Malý et al., 2018). The paper also proves the more quantitative estimate

$0$06

for continuous $0$07 (Malý et al., 2018). By contrast, lower Lipschitz extensions may fail even for compact $0$08, and the closedness of $0$09 is necessary (Malý et al., 2018).

A later metric-derivative theory distinguishes three pointwise quantities: $0$10 For continuous $0$11, $0$12 is $0$13-lower semicontinuous, $0$14 is $0$15-upper semicontinuous, and $0$16 is upper semicontinuous (Maslyuchenko et al., 25 Sep 2025). On an open or convex subset $0$17 of a normed space,

$0$18

where $0$19 denotes the upper Baire limit function (Maslyuchenko et al., 25 Sep 2025).

The same paper gives a space-level interpretation on convex subsets of normed spaces: for continuous $0$20,

$0$21

and moreover

$0$22

Thus the mapping $0$23 is an isometric injection of the Lipschitz space into the bounded functions on $0$24 (Maslyuchenko et al., 25 Sep 2025). This is not the same object as the Banach subspace $0$25, but it shows that little Lipschitz behavior can also be encoded as a pointwise differential invariant rather than as a closed subspace.

Finally, the ultrametric free-space theory adds a geometric extremal result: if $0$26 is ultrametric, then the unit ball of $0$27 contains a strongly extreme point, so $0$28 is not locally almost square (Abrahamsen et al., 26 Sep 2025). The construction uses ultrametric partitions and an alternating-sum function built from nearest-point approximations $0$29 (Abrahamsen et al., 26 Sep 2025). A plausible implication is that, in ultrametric settings, the little Lipschitz structure is embedded in a broader geometry that is markedly more rigid than in general metric spaces.