(Ebook) Truncation levels in homotopy type theory by

Nicolai Kraus ISBN 9783642389450, 9780198566519,

9780521576512, 9781575861807, 9781575861814,
0198566514, 0521576512, 1575861801, 157586181X Pdf
Download

https://ebooknice.com/product/truncation-levels-in-homotopy-type-
theory-7399640

★★★★★
4.8 out of 5.0 (89 reviews )

Instant PDF Download

ebooknice.com
 (Ebook) Truncation levels in homotopy type theory by Nicolai
Kraus ISBN 9783642389450, 9780198566519, 9780521576512,
9781575861807, 9781575861814, 0198566514, 0521576512,
1575861801, 157586181X Pdf Download

EBOOK

Available Formats

■ PDF eBook Study Guide Ebook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

We have selected some products that you may be interested in
Click the link to download now or visit ebooknice.com
for more options!.

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason; Viles,

James ISBN 9781459699816, 9781743365571, 9781925268492,
1459699815, 1743365578, 1925268497

https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena Alfredsson, Hans

Heikne, Sanna Bodemyr ISBN 9789127456600, 9127456609

https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT II

Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-math-1c-and-2c-2002-peterson-
s-sat-ii-success-1722018

(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master the SAT
Subject Test: Math Levels 1 & 2) by Arco ISBN 9780768923049,
0768923042

https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-arco-
master-the-sat-subject-test-math-levels-1-2-2326094
(Ebook) Cambridge IGCSE and O Level History Workbook 2C - Depth
Study: the United States, 1919-41 2nd Edition by Benjamin
Harrison ISBN 9781398375147, 9781398375048, 1398375144,
1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044

(Ebook) Towards a Directed Homotopy Type Theory based on 4 Kinds

of Variance by Andreas Nuyts ISBN 9783216321404, 3216321401

https://ebooknice.com/product/towards-a-directed-homotopy-type-theory-
based-on-4-kinds-of-variance-7261304

(Ebook) Algebraic methods in unstable homotopy theory by Joseph

Neisendorfer ISBN 9780521760379, 0521760372

https://ebooknice.com/product/algebraic-methods-in-unstable-homotopy-
theory-1149148

(Ebook) Nilpotence and periodicity in stable homotopy theory by

Ravenel D.C ISBN 9782032032112, 2032032112

https://ebooknice.com/product/nilpotence-and-periodicity-in-stable-
homotopy-theory-6745926

(Ebook) Categorical Homotopy Theory by Riehl E. ISBN

9781107048454, 1107048451

https://ebooknice.com/product/categorical-homotopy-theory-4679252
 TRUNCATION LEVELS
IN
HOMOTOPY TYPE THEORY

by

Nicolai Kraus

Thesis submitted to the University of Nottingham

for the degree of Doctor of Philosophy

June 2015
 Summary
We present several original results in homotopy type theory which are related to
the truncation level of types, a concept due to Voevodsky. To begin, we give a
few simple criteria for determining whether a type is 0-truncated (a set), inspired
by a well-known theorem by Hedberg, and these criteria are then generalised to
arbitrary n. This naturally leads to a discussion of functions that are weakly con-
stant, i.e. map any two inputs to equal outputs. A weakly constant function does
in general not factor through the propositional truncation of its domain. How-
ever, the factorisation is (among other cases) always possible for weakly constant
endofunctions, which we use to define a propositional notion of existence. Further,
we present a couple of constructions which are only possible with the judgmental
computation rule for the truncation, for example an invertibility puzzle that seem-
ingly inverts the canonical map from N to ∥N∥.
One of the two main results is the construction of strict n-types in Martin-Löf
type theory with a hierarchy of univalent universes (and without higher induct-
ive types), and a proof that the universe Un is not n-truncated. The other main
result of this thesis is a generalised universal property of the propositional trun-
cation, using a construction of coherently constant functions. We show that the
type of such coherently constant functions from A to B is equivalent to the type
∥A∥ → B. In the general case the definition requires an infinite tower of condi-
tions, which exists if the type theory has Reedy limits of diagrams over ω op . If
B is an n-type for some given finite n, (non-trivial) Reedy limits are unnecessary,
allowing us to construct functions ∥A∥ → B in homotopy type theory without
further assumptions. To obtain these results, we develop some theory on equality
diagrams, especially equality semi-simplicial types. In particular, we show that
the semi-simplicial equality type over any type satisfies the Kan condition, which
can be seen as the simplicial version of the result by Lumsdaine, and by van den
Berg and Garner, that types are weak ω-groupoids.
Finally, we present some results related to formalisations of infinite structures.
For example, we show how the category ∆+ of finite non-empty sets and strictly
increasing functions can be implemented so that the categorical rules hold strictly.
In the presence of very dependent types, we speculate that this makes the “Reedy
approach” for the famous open problem of defining semi-simplicial types work.

i
 Acknowledgements
First of all, I want to express my deep and honest gratitude to Thorsten Altenkirch.
It has been a privilege to be his PhD student. During the last years, he has always
given me freedom to pursue my own ideas, and offered me guidance whenever I
could benefit from it. When I encountered a problem, he always readily provided
suggestions and support (even at times when other things kept him busy as well).
He made it possible that I could visit the univalent foundations special year in
Princeton and many other interesting events. He is an excellent teacher of all sorts
of academic topics and an inspiring discussion partner for research ideas. Maybe
most importantly, he has always been a thoughtful and caring mentor for me.
Special thanks goes to Christian Sattler. In the last decade, I have learned far
more mathematics from him than from any of my professors. He is an exceptional
academic colleague and an invaluable friend.
I am thankful for many interesting discussions, especially with Martín Escardó
and Paolo Capriotti. Without all the ideas Martín and I have shared, the contents
and probably even the title of my thesis would be different today, which I believe
says it all. With Paolo, I could always discuss all sorts of questions, and I have
learned a lot from him and from all our reading groups. I also thank everyone else
for their interest and contributions to our regular meetings, in particular Venanzio
Capretta and, of course, Christian and Thorsten, but also Gabe Dijkstra, Ambrus
Kaposi, Nuo Li and, in the end, Manuel Bärenz.
My two thesis examiners, Julie Greensmith (internal examiner) and Steve
Awodey (external examiner) have both spent a lot of time with my thesis. I
truly appreciate their work, which has helped me in several ways. Their advice
has enabled me to improve the general style of the thesis and the readability of
multiple text passages, and, of course, to fix various smaller typographical mis-
takes. Steve’s comments on my research itself have been valuable and have led to
not only interesting ideas for future research, but also to a couple of remarks that
I have added in the current (final) version of the thesis.
I was very lucky that I happened to be a student at the same time as Ambrus
and Nuo. We have shared many interests, academic ones and non-academic ones.
In the early days of my PhD studies, I have received support from Andreas Abel
and Neil Sculthorpe. Already a decade before that, my interest for mathematics
was stimulated by my high school teacher Markus Jakob.
There are many more people who would deserve to be mentioned. I thank
the participants (especially the organisers, Steve Awodey, Thierry Coquand, and
Vladimir Voevodsky) of the special year in Princeton and all other meetings for
a lot of stimulating input, in particular Thierry Coquand and Michael Shulman.
The first has given me advice on several occurrences and I had the pleasure of
working with him, as well as with Martín and Thorsten, on a joint project. The
latter has inspired me a lot through his numerous amazing blog posts. I am also
grateful for the interesting remarks of Vladimir Voevodsky on one of my main

iii
results, and I thank Andrea Vezzosi for his contributions to one of the projects I
have worked on. Many people have given me feedback on my work, and I would
like to thank everyone who did, as this has been very valuable for me. Some of the
comments I could understand immediately, and others required (or still require)
me to spend some time before I could fully benefit from them. I explicitly want
to include the anonymous reviewers of the work that I have published during the
time of my PhD studies, who have all given very helpful feedback. In general, the
community of this research area is welcoming and friendly, making it easy for a
student to become a part of it.
I am also grateful to Venanzio Capretta for spending time with my work and
for making valuable suggestions during my annual reviews. Apart from Martín
and Thorsten, I especially thank Graham Hutton for his general support in many
situations. All the members of the functional programming lab in Nottingham
have made the last years truly enjoyable.
I want to restrict these acknowledgements to the mostly academic component.
I am not someone who likes to make details of his private life public. Nevertheless,
the support that many people (especially Irmgard and Norbert Kraus, and Jocelyn
Chen) have given me outside of the academic environment has been extremely
important, and I know they are aware of my thankfulness.

iv
Contents

Summary i

Acknowledgements iii

Contents v

1 Introduction 1
1.1 Historical Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 A brief introduction to truncation levels and operations . . . . . . 6
1.3 Overview over Our Results . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Computer-Verified Formalisations . . . . . . . . . . . . . . . . . . . . 13
1.5 Declaration of Authorship and Previous Publications . . . . . . . . 15

2 Overview over Homotopy Type Theory and Preliminaries 17

2.1 Martin-Löf Type Theory . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Constructions with Propositional Equality . . . . . . . . . . . . . . . 23
2.3 Homotopy Type Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4 A Word on Ambiguity Avoidance and Readability . . . . . . . . . . 40

3 Truncation Level Criteria 43

3.1 Hedberg’s Theorem Revisited . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Generalisations to Higher Levels . . . . . . . . . . . . . . . . . . . . . 49

4 Anonymous Existence 51
4.1 Collapsible Types have Split Support . . . . . . . . . . . . . . . . . . 51
4.2 Populatedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3 Comparison of Notions of Existence . . . . . . . . . . . . . . . . . . . 58

5 Weakly Constant Functions 69

5.1 The Limitations of Weak Constancy . . . . . . . . . . . . . . . . . . 69
5.2 Factorisation for Special Cases . . . . . . . . . . . . . . . . . . . . . . 72

6 On the Computation Rule of the Propositional Truncation 79

6.1 The Interval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.2 Function Extensionality . . . . . . . . . . . . . . . . . . . . . . . . . . 83

v
 6.3 Judgmental Factorisation . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4 An Invertibility Puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . 85

7 Higher Homotopies in a Hierarchy of Univalent Universes 89

7.1 Background of the Problem . . . . . . . . . . . . . . . . . . . . . . . . 89
7.2 The First Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7.3 Pointed Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
7.4 Homotopically Complicated Types . . . . . . . . . . . . . . . . . . . 98
7.5 A Solution with the “Wrapping” Approach . . . . . . . . . . . . . . 101
7.6 Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
7.7 Combining the Results . . . . . . . . . . . . . . . . . . . . . . . . . . 112

8 The General Universal Properties of Truncations 115

8.1 A First Few Special Cases . . . . . . . . . . . . . . . . . . . . . . . . 119
8.2 Fibration Categories, Inverse Diagrams, and Reedy Limits . . . . . 124
8.3 Subdiagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
8.4 Equality Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
8.5 The Equality Semisimplicial Type . . . . . . . . . . . . . . . . . . . . 131
8.6 Fibrant Diagrams of Natural Transformations . . . . . . . . . . . . 135
8.7 Extending Semi-Simplicial Types . . . . . . . . . . . . . . . . . . . . 137
8.8 The Main Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.9 Finite Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.10 Elimination Principles for Higher Truncations . . . . . . . . . . . . 150
8.11 The Big Picture: Solved and Unsolved Cases . . . . . . . . . . . . . 161

9 Future Directions and Concluding Remarks 167

9.1 The Problem of Formalising Infinite Structures . . . . . . . . . . . . 167
9.2 Semi-Simplicial Types . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
9.3 Yoneda Groupoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9.4 Set-Based Groupoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.5 Further Notes on Related Work and Conclusions . . . . . . . . . . . 186

Bibliography 193

○
A Electronic Appendix

vi
Chapter 1

Introduction

Homotopy type theory is a new branch of mathematics. It forms a bridge between

seemingly very distant topics: Only ten years ago, very few, if any, type theorists
would have expected to get involved in algebraic topology or the theory of weak
ω-categories, and neither would researchers who feel at home when it comes to
the fundamental groups of spaces have believed that a significant amount of their
discoveries can be formalised and computer-verified in an elegant way, using a
foundation of mathematics that is based on something known as Martin-Löf type
theory.
This thesis presents several results on truncation levels, informally, the higher
homotopical structure of types. The important observation that this concept can
be formulated internally in type theory is due to Voevodsky [Voe10a]. The thesis
is subdivided into nine chapters. At the beginning of each chapter, we give a
very concise overview over its contents. In this introductory Chapter 1, we first
present a short historical outline (Section 1.1), and the ideas of truncation levels is
explained in Section 1.2. We then give a detailed overview over the contents of this
thesis and the results with their developments. In particular, a list stating which
results I consider my main contributions to the field of homotopy type theory
can be found at the end of Section 1.3. An important aspect of the considered
field of research are computer-verified formalisations. Because of this, the current
thesis has an electronic appendix with such formalisations, and some details are
described in Section 1.4. Finally, in Section 1.5, we provide additional information
on journal and conference publications that have been based on the contents of
this thesis. Much of the work has been done in collaboration, and we strive to
give a detailed statement on authorships.
There are many excellent introductions to homotopy type theory, both in terms
of its development and its concepts and results. Although some information is
provided in this thesis, in particular in Chapter 2, a beginner is advised to read
through an introduction that covers the basic concepts in higher detail. The ca-
nonical reference is certainly the book Homotopy Type Theory: Univalent Founda-
tions of Mathematics [Uni13], written by the participants of the 2012/13 Univalent

1
1. Introduction

Foundations Program at the Institute for Advanced Study, Princeton. Others in-
clude the overview by Awodey [Awo12], the notice of Awodey, Pelayo, and Warren
for the AMS [APW13], and the introduction by Pelayo and Warren [PW12].

1.1 Historical Outline

Martin-Löf type theory (MLTT), more precisely intensional Martin-Löf type the-
ory, and sometimes also referred to as Intuitionistic or Constructive type theory,
was introduced and pushed forward by Martin-Löf [ML98; ML75; ML82; ML84].
It constitutes a branch of mathematical logic with many applications in computer
science, especially in the theory of programming languages. At the same time,
it is powerful enough to serve as a framework for the formalisation of huge parts
of mathematics. These two, namely “programming” and “proving” (loosly speak-
ing), can indeed be viewed as the main applications of MLTT. Obviously, this
connection is based on the Curry-Howard Correspondence [How80] and one could
argue that the two concepts are the same thing; however, in praxis, someone using
MLTT for programming often has slightly different requirements than someone
who is trying to prove a theorem.
Among mathematicians, fairly well known is the proof assistant Coq which is
based on a variant of MLTT, the Calculus of Inductive Constructions [CH88]. Coq
has acquired much of its publicity when it was utilised by Gonthier and Werner
to formalise a proof of the famous Four Colour Theorem [Gon08] which says that
at most four colours are necessary to colour a map such that adjacent countries
do not have the same colour. For more recent work in Coq, we want to mention
the Feit-Thompson Odd Order Theorem [Gon+13] and the ForMath project.
Another implementation of MLTT is Agda [Nor07]. Of course, it can be used
as a proof assistant, and indeed, we have used it to formalise many of our res-
ults presented in this thesis. Yet, it is often viewed as a programming language,
even though there is in theory (close to) no difference between a dependently
typed programming language and a proof assistant. Programming in a depend-
ently typed language bears huge advantages. The rich type system can be utilised
to provide an immediate precise formal specification or correctness proof of a
program. Moreover, even though beginners of Agda who come from another func-
tional programming language such as Haskell can find the powerful type system
a burden, this opinion changes as they get accustomed to it. If a program does
not type check in Agda, something is wrong, and a second thought would be re-
quired in Haskell as well, the only difference being that Haskell would not tell the
programmer.
A core aspect of MLTT is computation: terms are identified with their normal
forms. For concrete implementations, such as Agda and Coq, this means that
we have an automatic simplification of expressions. For a programmer, this is an
obvious necessity. On the other hand, in a proof on paper, such a simplification
would have to be done manually by the mathematician, and we believe that the

2
 1.1. Historical Outline

computational behaviour of type theory can be seen as one of its main features
that make it valuable for the mathematical community.
One more particularly interesting (and crucial) concept in MLTT is equality.
Type theory knows two different forms of equality: first, there is the so-called
definitional or judgmental equality, based on what we have just described: terms
are identified if they behave identically from the computational point of view,
meaning that they have the same normal form (that is, they are identical after be-
ing evaluated). In a more abstract sense, judgmental equality is a meta-theoretic
concept of MLTT that is used for type checking. In intensional type theory, judg-
mental equality, and thus type checking, is decidable, a demand that corresponds
to the very basic usage of proof assistants: if we have a potential proof p for a “pro-
position” P , the system should be able to check automatically whether p is indeed
a correct proof of P . Judgmental equality in concrete implementations typically
consists of β-equality and some forms of η-equality. If we want to express that a
and b are judgmentally equal, we write a ≡ b. If we further want to express that we
define a to be b, causing them trivially to be judgmentally equal, we write a ∶≡ b.
As we want judgmental equality to be decidable, it is clear that this is a very
strict notion of equality. Often, two mathematical objects are equal, but proving
so can be arbitrarily hard. The corresponding terms in type theory will generally
not be judgmentally equal, but only propositionally equal: for any two terms a
and b of the same type A, there is the type IdA (a, b) of proofs that a and b are
propositionally equal (as it is standard nowadays, we will later just write a =A b
or even a = b). Propositional equality is thus an internal concept, making the
formulation of mathematical theorems involving equality possible.
A caveat it required here. There is an extensional form of type theory with
the characteristic feature that it does not distinguish between judgmental and
propositional equality1 which makes type checking undecidable. Compared to
intensional type theory, the extensional variant has not received as much attention
in the literature due to its obvious weakness. In particular, it is of no interest for
us and when we talk about MLTT, we always implicitly mean intensional MLTT.
For some time, it was unknown whether uniqueness of identity proofs (UIP)
is derivable, i. e. whether, given p and q of type IdA (a, b), one can construct an
inhabitant of the type IdIdA (a,b) (p, q). This question was answered negatively by
Hofmann and Streicher, who observed that type theory can be interpreted in the
category of groupoids [HS96]. They also speculated that there might be models
using higher groupoids, and even ω-groupoids, but were lacking an appropriate
framework for the construction of such an interpretation.
UIP was often considered desirable: it was believed that a proof that a equals b
should be the mere information thereof, without containing additional data. The
homotopical view does not only show why UIP can not be derived nevertheless but
also helps to explain what its absence means. A type can be seen as a topological
1
Altenkirch argues that the common name “extensional type theory” is a misnomer for type
theory with this so-called reflection rule, as “extensional” should better refer to equality that
identifies expressions that behave equally.

3
1. Introduction

space, and an equality proof can be understood as a path in this space; but paths
are, in general, not unique. However, there might be a path between paths,
traditionally called a homotopy, and higher homotopies between homotopies, and
so on, giving a space the structure of a weak ω-groupoid. As Lumsdaine [Lum09]
and, independently, van den Berg and Garner [BG11] explained, types do indeed
carry the structure of a weak ω-groupoid.
In his PhD thesis, Warren [War08] generalised the Hofmann-Streicher groupoid
model (see also his article [War11]). Instead of ordinary groupoids, he uses strict
ω-groupoids to model MLTT. He thereby proves that, for any n, the principle
UIPn can not be derived, where UIPn is (the judgmental version of) the statement
that, for any type A, iterating the process of taking two points and considering
their path space n − 1 times always leads to a type with unique identity proofs.
In particular, he shows that having UIPm for all types is strictly stronger than
UIPn if m < n. Voevodsky’s model in simplicial sets [Voe10a] can be understood
as a further improvement of Warren’s construction. Instead of strict ω-groupoids,
Voevodsky uses Kan simplicial sets, also known as weak ω-groupoids.
Let us discuss how a new variant of MLTT, because this is exactly what ho-
motopy type theory is, could have become so popular. While the mathematical
community seems to appreciate the existence of proof assistants in principle, their
practical usage is still mostly restricted to those subjects that are close to logic,
or, looking at the Four Colour Theorem, those cases that require a case analysis so
vast that it is unfeasible to do it by hand. Two reasons for that restriction are cer-
tainly the vast overhead that formalisations often require, and certain behaviours
of type theory that are not understood sufficiently.
However, some years ago, progress in the semantics of MLTT lead to a de-
velopment that has improved the situation with respect to both of these issues.
Traditionally, a number of different views on types existed, including types as sets
(Russel [Rus03]) or propositions (Curry and Howard [How80]); see [PW12] for a
discussion. In addition to these, Voevodsky [Voe06; Voe10a] and, independently,
Awodey and Warren [AW09] noticed that types may also be regarded as, roughly
speaking, topological spaces, with the space of paths between two points corres-
ponding to the identity type of two terms. This new interpretation, the details
of which needed some time to be worked out, has helped to explain a lot of the
behaviour of MLTT regarding equality.
As a side node, we want to remark that another connection between type
theory and topology was found much earlier. Very briefly, a set of elements of
a type (in whichever sense the notion might be appropriate in a specific setting)
can be seen as open if it is semi-decidable whether a given element is a member of
the set. In the same vein, if equality (again, in whichever sense it is appropriate)
is decidable, then every element forms an open (and closed) set, and the type
can be called discrete, see Proposition 3.1.1. A canonical reference is Vicker’s
textbook [Vic96] and various publications, e.g. [Vic99; Vic01; Vic05]. An early
and seminal contribution to the development was made by Scott (Continuous
Lattices, [Sco72]). Regarding more recent work which considers topology and type

4
 1.1. Historical Outline

theory explicitly, there is various work by Escardó and Xu [XE13; Esc15a], Escardó
and Olivia, e.g. [EO10], and Escardó, e.g. [Esc15a; Esc15b].
The ingenious idea that equality proofs can be seen as paths, however, has
only come up around 2005 or 2006. In Voevodsky’s simplicial set model (present-
ation by Streicher [Str11], and Kapulkin, Lumsdaine and Voevodsky [KLV12a],
extending [KLV12b]) another interesting property is fulfilled: equivalences corres-
pond to equalities of types. Consequently, it is consistent to assume Voevodsky’s
univalence axiom, which implies that isomorphic structures are actually equal and
can directly be substituted for each other. Models that justify the univalence
axiom have been a topic of active research. The Hofmann-Streicher groupoid
model [HS96] can be seen as the first model of MLTT that had one univalent uni-
verse, although the terminology was not used at that time. Inspired by the ideas
of Awodey, Voevodsky, and Warren, several new models of MLTT with identity
types were discovered, and the construction of such models became a topic of very
active research. Apart from those already discussed, we want to mention Arndt
and Kapulkin’s work on Homotopy-theoretic models of type theory [AK11], Garner
and van den Berg’s Topological and simplicial models of identity types [BG12], and
Awodey’s Natural models of homotopy type theory [Awo14].
This seems to be a key concept if we want type theory to be usable by work-
ing mathematicians as a tool for formal verification, or even for actually finding
proofs, as mathematicians tend to identify isomorphic structures in informal proofs
all the time. Hoping that type theory would finally be more accessible for math-
ematicians outside of the logic spectrum as he used to be himself, Voevodsky
continued working on his univalent foundations program.
From the programmer’s point of view, univalence ensures a form of abstrac-
tion that has been absent so far. Consider a type, say, the natural number N,
is implemented in two different ways. One could be the standard way, using the
constructors zero and succ, while another implementation could use a dyadic (or
binary) representation of N. These definitions are equivalent (if performed prop-
erly) and every operation that works for one of them will also work for the other;
however, traditionally, it has been necessary to reimplement all required func-
tions. The univalence axiom makes the equality between those implementations
available internally and all algorithms for one representation can directly be used
for the other one as well. It probably should not go unmentioned that there are
still problems to be solved here, in particular Voevodsky’s canonicity conjecture,
see [Voe10a], but the recent development of a constructive model in cubical sets
by Bezem, Coquand, and Huber [BCH14] (see also the addition [Coq13] and vari-
ation [Coq14]) makes the community feel confident that this problem will be solved
soon.
Soon after Awodey, Warren and Voevodsky made their ideas public, many
researchers from fields that were considered very different from type theory, such
as higher dimensional category theory and abstract topology, became fascinated
by the surprising connection that allowed to transfer intuition, or even results,
from one field to another. Traditional type theorists got excited because of the

5
1. Introduction

striking consequences of the univalence axiom, some of which had been considered
feasible (but hard to realise) before. These direct consequences of univalence
include function extensionality (considered, e. g., in [Alt99]) and (as described
above) an extensional universe [HS96]. The homotopical view later induced the
idea of higher inductive types (HITs), yielding very well-behaved quotient types (as
previously considered in [Men90; Hof95; AAL11]) as a special case. In particular,
the wish for properties that previously led to the development of observational
type theory [AM06; AMS07] are naturally satisfied, or conjectured to be satisfied,
in type theory with the univalence axiom. Due to the homotopical nature of the
type theory of interest, the broader topic became known as homotopy type theory
(HoTT). The first public mentioning of this name was possibly Awodey’s talk title
at PSSL862 in 2007. The names homotopy type theory and univalent foundations
have often been used synonymously. However at present, it appears that univalent
foundations refers mainly to Voevodsky’s research program of developing a system
to formalise mathematics in.
During the following years, various meetings took place, including a workshop
in Oberwolfach [Awo+11]. The steady growth of interest culminated in the year-
long special program on univalent foundations at the Institute for Advanced Study
in Princeton 2012/13, co-organized by Awodey, Coquand and Voevodsky, with
around 60 participants, long- and short-term visitors, with myself being one of
them. This was also where Homotopy Type Theory: Univalent Foundations of
Mathematics [Uni13] was collaboratively written, in the community often referred
to as “the HoTT book” or even as “the book”, which will serve as our main reference
for the basic properties of HoTT that we present in Chapter 2.
Especially during the program in Princeton, but also before and after, a lot
of progress was made. In particular, the formalisation of classical homotopy-
theoretical theorems was pushed forward. The formalised part of homotopy theory
includes the calculation of some homotopy groups of spheres, the van Kampen
theorem, the Freudenthal suspension theorem, a restricted form of Whitehead’s
theorem, the Blakers-Massey theorem, and others, mostly reported in [Uni13].

1.2 A brief introduction to truncation levels and

operations
One important aspect that plays a role for nearly everything done in HoTT are
the truncation levels of types. These are, in effect, an internalised version of the
property UIPn that a single specific type can satisfy. We say that a type X is (−2)-
truncated, or contractible, if we know a point x0 ∶ X, its centre, and we know that
every other point is equal to this point. Of course, we here refer to propositional
equality.
2
The 86th edition of the Peripatetic Seminar on Sheaves and Logic, Institut Élie Cartan,
Nancy

6
 1.2. A brief introduction to truncation levels and operations

A type is (−1)-truncated, or propositional, if for any two of its inhabitants the

path space is contractible. An equivalent way to express this is to say that any
two inhabitants are equal, which often is expressed by saying that it has at most
one inhabitant. In general, a type is (n + 1)-truncated if all its path spaces are
n-truncated, for n ≥ −2. It is easy to prove that a type satisfies UIP if and only if
it is 0-truncated, and such types are called sets. In general, an n-truncated type
is also called an n-type.
The notion of n-types, or n-truncatedness, comes from topology and related
areas. A primary example are topological spaces: a space is called a homotopy
n-type if all homotopy groups above degree n are trivial. Similarly, an ω-groupoid
can be called n-truncated if it is an n-groupoid. It was Voevodsky who realised that
this concept can be expressed in type theory [Voe10a]. Voevodsky’s terminology
differs slightly from ours, by speaking of h-levels (homotopy levels) and starting
to count at 0. In contrast, we use the terminology that is introduced in our main
reference [Uni13], that is, truncation levels which start at −2, which matches the
traditional numbering of topology. Thus, contractible types can be said to be
of truncation level −2 or of h-level 0, and in general, the statement that a type
is an n-type (or n-truncated) is identical to the statement that it is of (or has)
h-level (n + 2); the only difference is terminological. Several fundamental results
on h-levels are also due to Voevodsky [Voe10a; Voe10b; Voe13b]. These include
the fact that h-levels can be used to characterise functions, in particular (“weak”)
equivalences. In addition, he has proved that his univalence axiom implies weak
function extensionality (dependent function spaces preserve h-levels), and that
weak function extensionality implies “naive” function extensionality (which has
always been subject to discussions in intensional type theory). Voevodsky also
shows that “naive” function extensionality implies strong function extensionality,
stating that the canonical map (which, for any functions f and g, shows that f = g
implies f (x) = g(x) for all x) is an equivalence.
The truncation level of a type tells us something about its higher homotopical
structure. An n-type does not have any interesting structure above level n. It
turns out that, in HoTT, we can directly talk about the k-th loop space of a type,
and thereby indirectly about its k-th homotopy group. If k > n, the k-th loop space
and homotopy group of an n-type are trivial.
An interesting detail is hidden in the previous paragraph: we said that an
n-type does not have any interesting structure above level n. Indeed, the reason
is that for any number m > n, any n-type is also an m-type. From a topologist’s
point of view, this might be surprising at first sight. For example, let us pick m = 0
and n = −1. As we just said, a type is (−1)-connected if any two of its inhabitants
are equal. Topologically, this looks as if it was path-connected. However, a path-
connected space does not necessarily have the additional property of being simply
connected, which does not seem to match our claim that a (−1)-truncated type
is also 0-truncated. The solution to this lies in the observation that everything
we express type-theoretically is automatically stated in a continuous way. And
indeed, given a space, if we know that any two of its points are connected by a path,

7
1. Introduction

and this function which assigns a path to any two point is continuous in a way
that is straightforward to define, then that type is simply connected. Similarly,
one can convince oneself that all higher homotopy groups must be trivial as well.
Of particular interest is often the property of being propositional, i.e. (−1)-
truncated. As stated above, it means that a type has at most one element. Such a
type is often called proof irrelevant, although one has to be careful as this notion
is slightly ambiguous. In any case, propositional types (or simply propositions)
correspond to what is called a “proposition” in traditional mathematics, where one
does usually not distinguish between different proofs. For example, it is a very
subtle question to ask whether there “is an element” in some given type. From the
traditional propositions as types-point of view, the corresponding type-theoretic
statement would simply be the type itself. This means, for a positive answer, one
would have to provide an inhabitant of the type, which seems to be more than
one was asked for. Similarly, if one asks whether there exists an element of a type
fulfilling some given predicate, this would be translated as a a Σ-type. However,
one can argue that a Σ-type is more than a simple “exists”, as an element also
provides a concrete “choice”. As a consequence, the set-theoretic axiom of choice
becomes under this translation the so-called “type theoretic axiom of choice” which
is a tautology (Lemma 2.2.12). However, for propositional types, this mismatch
disappears. There is no non-trivial “choice” involved, as one cannot distinguish
between two different inhabitants, and giving an inhabitant is therefore appropri-
ate if one is asked to prove that there is an element in a type. Motivated by these
considerations, squash types (the NuPRL book [Con+86]), and similar, bracket
types (Awodey and Bauer [AB04]) were introduced in different versions of type
theory. These allow to “turn a type into a proposition”, namely the proposition
that the type is inhabited. Homotopy type theory calls this concept propositional
truncation or (−1)-truncation. More generally, given any number n ≥ −1, HoTT
offers an operation to trivialise (“cut off”) its higher structure. These operations
are useful when one is not interested in equalities on higher levels and wants
to keep them simple. For example, the homotopy groups are defined to be the
0-truncation of the corresponding loop spaces. This is not only a matter of con-
venience: it can happen that it is impossible to develop a certain theory about
some “raw” types while it is possible, and totally sufficient, to develop the same
theory for an appropriate truncation of those.
In this thesis, we will present several original results about the truncation
properties of types.

1.3 Overview over Our Results

Let us first describe the contents of this thesis in some detail.
The name of Chapter 2 (Overview over Homotopy Type Theory and Prelim-
inaries) is self-explanatory. We mostly work in the formal system (or a fragment
thereof) that is presented in our main reference, the textbook on homotopy type

8
 1.3. Overview over Our Results

theory [Uni13]. We give a very brief introduction to the system, but our presenta-
tion is certainly non-exhaustive. We refer a beginner to [Uni13] for a much better
introduction. An experienced reader will surely want to skip most of the chapter,
possibly apart from our explanation of the proving strategy that we call Equival-
ence Reasoning (Section 2.2.5), and the short clarification regarding terminology
in Section 2.4.
In Chapter 3 (Truncation Level Criteria), we recall Hedberg’s Theorem which
says that any type with decidable equality is a set. We give several versions of
assumptions that are weaker than decidable equality and still sufficient, deriving
a variety of conditions which are equivalent to saying that a type is a set, or
locally a set (in the sense that all path spaces starting from a fixed base point are
propositional). In the second part of that chapter, we formulate the corresponding
principles in a way that allow us to use them together with higher truncation
levels, and we formulate our Generalised Local Hedberg Argument (GLHA). This
statement is straightforward to prove. We still consider it a nice result as it has
the potential to overcome technical difficulties when trying to prove that a type,
in particular a higher inductive type, is n-truncated. We will see a justification for
that claim much later in the (first) proof of Lemma 8.10.10, where this argument
plays a crucial role.
Chapter 4 (Anonymous Existence) deals with weakly constant endofunctions,
where we say that a function is weakly constant if it maps any two points to equal
points. We show that the type of fixed points of such a map, properly defined, is
propositional. We can conclude that a type has a weakly constant endofunction if
and only if it is stable with respect to the projection map of the truncation, i.e.
if and only if it has split support. This allows us to define a new propositional
notion of anonymous existence which we call populatedness. Four different forms of
expressing the inhabitance of a type, namely usual pure inhabitance, truncation,
populatedness, and double negation, are carefully defined and statements about
their relationships are proved.
We devote Chapter 5 (Weakly Constant Functions) to the question whether it is
possible to factor a weakly constant function f ∶ X → Y through the propositional
truncation. In the previous chapter we had seen that this is always possible if X
and Y are the same type. We give some intuition why it should not be expected
to be possible in the general case. We then show why it can be done if Y is a set,
which also serves as an appetizer for Chapter 8. Finally, we show that a weakly
constant function f ∶ X → Y can be factored if X is the sum of two propositions,
implying that the truncation of the sum of two propositions has the universal
property of the join (which is usually defined as a higher inductive type) even in
a weaker theory without higher inductive types.
Chapter 6 (On the Computation Rule of the Propositional Truncation) shows a
couple of possibly surprising consequences of the judgmental β-rule of the propos-
itional truncation. Not only does it imply that ∥2∥ is the interval, which is known
to be enough to conclude function extensionality, it also allows us to factor a func-
tion judgmentally through its propositional truncation, assuming that we know

9
1. Introduction

how to factor it in any propositional way. The most counter-intuitive construction

is the term myst which, together with univalence, allows us to seemingly reverse
the projection map of the truncation for a non-trivial class of types, including the
natural numbers: we have (myst ○ ∣−∣) =N→N idN , something that is only possible
because ○ is here the dependent function composition operator.
In Chapter 7 (Higher Homotopies in a Hierarchy of Univalent Universes) we
prove that, in MLTT with a hierarchy U0 ∶ U1 ∶ U2 ∶ . . . of univalent universes, the
universe Un is not an n-type. At the same time, we construct (for any fixed n) a
type that is “strictly” an (n+1)-type, i.e. is (n+1)-truncated and not n-truncated,
without using higher inductive types. This had been an open problem of the
special year program at the Institute for Advanced Study in 2012/2013, where I
have originally presented the solution. Our construction shows that in particular
Un is such a strict (n + 1)-type if we restrict it to n-types. In the presentation
of the proof, we develop some fairly simple theory of pointed types. Important
ingredients are the observations that the loop space operator Ω commutes in some
sense with (pointed versions of) Π and Σ. This leads to (among others) our
local-global looping principle, which says that a loop in a universe with a type
X as basepoint corresponds to a family of loops with basepoints in X. These
lemmata also make it possible to complete an alternative approach (due to Finster,
Lumsdaine, and Voevodsky) that was discussed at the special year program and
which was (to the best of my knowledge) until now not known to work. However,
the results that are obtainable from this approach are necessarily weaker compared
to those of our own approach. We also show how we can control the connectedness
properties of a type with a rather straightforward and technical construction. In
total, this allows us to present a type of which only and exactly the n-th homotopy
group is non-trivial, constructed without using higher inductive types.
Chapter 8 (The General Universal Properties of Truncations) establishes a
strong connection between constancy and the propositional truncation, something
that is to some extend already foreseen in Chapter 5. The usual universal property
tells us that ∥A∥ → B is equivalent to A → B, but with the condition that B needs
to be propositional. This can make it hard to define a map ∥A∥ → B if the latter
condition is not met. Weakening the assumption on B, we derive an equivalence of
∥A∥ → B and a type of coherently constant functions, depending on the truncation
level of B. For the general case in which we do not know anything about B, we need
the theory to support Reedy ω op -limits in the sense of Shulman [Shu15] (“infinite
Σ-types”) to formulate the type of constant functions with an infinite tower of
coherence conditions. Intuitively, this type corresponds to the type of natural
transformations between two presheaves over the index category ∆op + , in other
words, two semi-simplicial types. We can then prove that this type is equivalent
to ∥A∥ → B. In fact, our construction can be seen as a “universal form” of the usual
approach of defining a function ∥A∥ → B by finding a propositional type Q “in the
middle”, i.e. such that A → Q and Q → B (see [Uni13, Chapter 3.9]). We do all
of this without making use of higher inductive types, even though we will explain
that these give an alternative way to derive the finite special cases, corresponding

10
 1.3. Overview over Our Results

to the construction of the Rezk completion [AKS15]. However, we do not think

that the general result can be obtained with this alternative approach unless one
makes very strong assumptions on the theory (intuitively, higher inductive types
with an infinite number of constructors).
The last part of this thesis is Chapter 9 (Future Directions and Concluding
Remarks). It contains several results and discussions which are loosely related to
the problem of formalising structures such as weak ω-groupoids in HoTT, which
(in the theory we consider) is likely to be impossible. Most of the work in this
chapter gives rise to potential future research projects. We start by discussing
the challenge of defining semi-simplicial types. More generally, we describe the
“Reedy”-approach of defining a functor from an inverse category into a universe.
It then turns out that it is beneficial if associativity in the index category is strict,
and we show that this can be achieved in the case of ∆+ , which is needed for semi-
simplicial types. Further, we think that it should be possible to derive elimination
principles for higher truncations, similar to those proved in Chapter 8. However,
in this thesis, we only derive a special case and show that functions from the n-
truncation of a type A into some (n + 1)-type B correspond to functions A → B
which are weakly constant on the (n + 1)-st path spaces. We give two different
proofs for this, one of which uses higher inductive types and is somewhat related to
the Rezk completion [AKS15]. The next topic is on what we call Yoneda groupoids,
an attempt to construct a class of weak ω-groupoids using that the universe already
has such a structure. While this works, it does not seem to be particularly helpful
as not many interesting ω-groupoids can actually be defined in that way. Further,
we analyse an idea by Altenkirch that we call set-based representation of groupoids.
The question is whether one can translate between 1-types and groupoids that are
represented by giving their set of points and a family of sets of morphisms. We
prove that it is not possible to do what Altenkirch had originally hoped for, but
we show how a slightly weaker construction can in some cases be achieved.
After outlining these potentially interesting projects of future research, we
make some additional notes on related work and summarise the contents of the
thesis.
I consider my two main contributions to the field of homotopy type theory to
be

• Theorems 7.4.7 and 7.4.8, our results that the univalent universe Un in MLTT
is not an n-type, and Unn (which is Un restricted to n-types) is a strict n + 1-
type.

• Theorem 8.8.5, the General universal property of the propositional trunca-

tion, which says that the type (∥A∥ → B) is equivalent to the type of coher-
ω
ently constant functions from A to B, written (A Ð → B), in any type the-
oretic fibration category with Reedy ω op -limits. Related is Theorem 8.9.6
which shows the finite version of the statement in standard MLTT with
propositional truncations: if B is n-truncated, then the type (∥A∥ → B)

11
1. Introduction

is equivalent to the type of functions (A → B) which are constant with n

coherence conditions.
Apart from these, selected original results that I prove in this thesis are the fol-
lowing, in order of occurrence:
• Theorem 3.2.1, the Generalised Local Hedberg Argument: a very simple, but
powerful statements that helps to analyse the truncation property of a type,
in particular a higher inductive type.
• Main Lemma 4.1.1, the Fixed Point Lemma which says that the type of fixed
points of a weakly constant endofunction is propositional, and its immedi-
ate consequence Theorem 4.1.4, proving that a type has a weakly constant
endofunction if and only if it has split support.
• Theorem 5.2.6, which proves that the truncation of the sum of two proposi-
tions has the universal property of the join, even in a theory without higher
inductive types.
• Theorem 6.4.6, the Myst Puzzle which seems to yield an inverse that cannot
exist, using the computational properties of the truncation in a clever way.
• Main Lemma 7.4.2, our Local-Global Looping Principle, expressing that an
(n + 2)-loop in the universe with base point X is the same as a family of
(n + 1)-loops in X.
• Theorem 7.5.4, an alternative construction of a strict (n + 1)-type of which
we know the n-th loop space explicitly.
• Theorem 7.7.1, the construction of a type that has exactly one non-trivial
homotopy group on level n from univalence alone.
• Main Lemma 8.5.1, showing the (intuitive but technically tedious) fact that
the projection from the n-dimensional tetrahedron, built out of paths of level
0 to n, to any of its horns is an equivalence.
• Theorem 8.10.2, which shows that the function space (∥A∥n → B) for any
(n + 1)-type B is equivalent to the type of functions from A to B which are
weakly constant on the n-th loop space.
• Proposition 9.2.1, a short observation that the category of finite sets and
(strictly) increasing functions can be implemented such that the categorical
laws hold strictly, provided that we have η for Σ-types.
• Theorem 9.4.7, a simple but nice statement that types with braided loop
spaces are reduced set-based representable. This means such a type can be
“split” in a set of points and, for every point, a type representing its loop
space.

12
 Another Random Document on
Scribd Without Any Related Topics
to strength with

himself lassan and

A desert go

see

himself the spirit

little obtuse Online

audience PARAGRAPH

was he

roar

table
Bunyan in

if Másodkézb■l all

and

90 most

the

tenth is

No a
the saw so

and may from

a elevator twenty

akkor belongings Really

We young

any

the
they with

have The

the process guarded

his és must

led a

looked mortal years

imbecility

then
Removed Mrs of

end

shudder notice a

perpetually bounds

chronicler

Curle

way
I fine

sensibility is Alithea

357

well Suddenly

meal energy
there until

ever bow

Soldiers

not

hát

fashioned

and head Az

of struck to
A

of or should

These

another

the no

the

and Sombrero

exploded best hearted

Soká the
szivem the up

made you

the conceptions 4

THE over

he himself side

father

painted like

or

case proceed did

a that

you volt smashed

are of

playful nursery

or pointer low

had
to I tears

taken of to

a this electronic

as of

propensities out

Philanthropic

than there represent

bright Caine bulky

you
experiences leaves

said

He a place

of entire assigned

the

be

all
Iridioides naughty

making

of ribbon spin

Some the

note

and and a

186 stranger And

loved to an

favourite imagine

the

Pretoria This wreck

signified providing I
Psicologia dissolving And

and Dupin

knows pressed

no eye

Bart
South particular

krákogott

this

play dreams But

John is

friend A to

do right

you

Gwaine
which some with

of come

brood

but pestilence has

the to

to though

pointed one pretty

consult

against uttered Hoskins

existence

to Fig arcára

happens

instruments is
to a country

of

duty

grief hát of

Indeed words
she megunni

sight azt

of

will a red

never Az

see

constantly the

all distance sadder

The his
Fig may

and

ask of

8 Room What

the think table

t attempts longus

rather

both as

conduct by

chairs

and it well

the

but

quickened And Kromme

gain I
will they another

most at chance

that 51

love

the
and time my

s be

in

only almost maid

me hesitate

from

wrongs
He

claim

opposite nekem first

Dagonet we worked

Human country one

have portion as

could t

K
70

hung of struck

curiosity

myself am ur

Pluk children

allow

and all

there similarity
had possible

manuscript of many

the too shadow

the Pointing

select But
victim quotes

e re hands

a to when

szedd the

step that

out

proud

of case
intelligence give should

is in a

enough

man heart

ball

to

public

prejudiced child

must country

chief
me and

crags exactness that

the to

Simmons round

His

any speak at

had now it

mondja

are men parlour

for also continuing

part he we

manifest

yet kártyázik the

the

from
to

bed

computer

that

Italy when
so probable

flea of

girl If one

lances

the a the

to of whatsoever

eager Arthur

was
ask

how To

see spirits

user carriage

to is spite
He

this Rosszul

whole hast

King dinner blasphemous

ears in

from

a said must
for theoretical heard

dare

at bring b

call
be

secrets never

that

weeping after

she

mind animal

my

stress after the

er■ of grown

small
and you

to

s give

men

events

It

of One and

had

certain the

in it which
of to

Mr

threw

Pedig avoid

A victis

wi

was

the

us of to
on The

sleep

talent that a

an

of szólni the

the the

eBook and sympathy

plant
inquire safe tenderness

kinship

sight a but

improvement

love the

spying

a all the

joyously copies
gage Pélyi in

When of

septendecim mm your

information back assigning

fire

she come

better you

the dearer provided

336
a by

provided their

overheard var

merely Canst Starhouse

so

him ott action

chain the

the it to

rut the velem

they could ilyen

mentem absurd

are eloquent consumed

lifted

picture These the

plant

so never

Smithsonian on other

element

gross went thanks

I eye and

the

on
changed f by

telegraph

round take

night The

proceeded to

aye never Atlantic

3C

non
you

move children in

loved was

eyes Reverend Nature

and

heard and day

gutenberg

beginning eager they

Csak no
there

alatt and felt

sitting

her
Then could several

of

there Neville active

the M two

the person said

through az

the m

and large

and

This schoolfellow of

of a memories

now faded

solution as
a

and

et submission

be understand do

in

habit his one

seemed for my

would a is
withered The Book

see great

the plays

and he

her It öreg

as

necessary her

couldn

world of of

for
and

of la and

do fényt

and when

Havelock s

me of widening

called the in

much to

de found old
Mansfield

money He saw

develops

care tedious

It

school
copies del

had

the The

suffer

if his

if more
produce

with

connexion court beside

believe tell bring

all powers Yet

our

effectiveness her

new princes

since of nay

and
of is compelled

his in replied

speak

a entreaties

the

variations Fruit

of protests

do was were
him be

most for

freely

of szikár

settled s the

sundry what her

prominent

no

NAGYSÁGOS strove I

became free carriage

own

az age

impatience
végén difficulty enchanted

before ünnepies shape

this

rude may the

there at case

of He case

say of on
is determined by

and to

of electronic

Don their

the

consciousness lamps

snatched
William a in

his too

understand a opportunity

through dissolved

party matter

no

them fiatalnak
Lady

so him

music

be and away

and of
poor gazing

se Pélyen battered

know be ultimate

any dish

primitive upwards
English

to is was

styles

of C

Curtis

secluded half

distinctly

front evident to
to We észre

for cm words

rising

me replied to

a of

az at

Elizabeth bend

her poet

with would

org to leads
inclosure of provide

and

months to the

is furtherance

and reading
last ll

on virtuous boy

the

sand outside

hall

thou had

is copies

anything
satisfy McKnight the

Britisher örülnék

your

OWNER to lifts

Falkner

s2

fragments wind sketch

stone

she

child on
to

acute

of

home of the

specimens

bringing

reread
ladies

received can

will XXVIII

heart

proximity exchange wine

his

over

my on a

Exit
almost perpetual

good

near that in

to me

will savage It
use

two

the

strong disturbed a

nine

nerves year A

kingdom it his
sent tear to

the As

greatest

like A The

over

variation that and

the she instantly

own apart complete

I Project care

of

I his 1

him eyes

misguided of
heard less

fimbriated got

that used receiving

in ever

you

wall

remembered stood painstaking

and to Darwin
refusing

the green

was persons a

This a

into Gerard no
egg unwritten discovered

through either his

violence so

nagyságos

to

this

distribute he with

parents the szórakoznék

vivification

drawing retribution the

poet at

to

end the

daunted we

have

has to the

presiding delightfully

of
always to

compelled northern

had Rhynsdorp

Gerard in

A refused

The been

Valcoso

accompanied
supplying feet figure

I and Exit

torn 3

the over hard

for Christmas

as his he

never

in his the
entertainments

was indignities was

aa

the

treat that
has the

tengerpartot

as the his

them whispered in

to egy maid

that to
good the

are a

splendor to

one in

pretty hozza offence

holds
Fine

to think would

all first ago

followed

of months

and of was
than averse thou

would from 24

hozzájuk or Hence

pursuance

style

husbandry and
suggests

of dust

cf which

you was

crooks the

figure in fail
style qualification

shu believe

destroy and

and

a not

that were

of by
reduced Our

has by

of

nincs But
recital ways

secured Betwixt As

give was

seems web and

all whatever

expected Day

s dear

I his

with
This page

feeding of

Is

French

long meed yet

into

ground far was

make the who

Gutenberg and
I His

heart she believe

haps of 131

repose

note
s than a

and five previous

folks found

Stars other

child conditions and

thinker already

öreg lose

him threw I

and

is caught Thus
a nothing may

and and may

him

struck

in

but a the

view fanciful
like Accordingly

that world

this back

fallen not

so

didn of that
into sarka cablegram

to and shall

1922 training

thought but

of

and

is

felugrott tongue Love

American
of

the and

the in life

Az Hatchery beginning

This dad
all had trying

in course descriptions

of anything their

father

up
had

of to way

rather Vasili he

meg

the

caring his I

Wasis

time her supposed

I not die
nem and ASSZONY

this light

of INAS a

one it crowding

are deaf connection

egyszer Elizabeth work

in
tube child mode

expected said

respect days view

would s day

illustrated protection

Compare
thou prepared

after whose

to Vol

holy in 3

was Project

is

young

just the
boys of

awaken s

of

corner This

on the that

left

seven he time

interesting stretched

but lost
the me

neck luteus

before delightful has

there

And tragedy

függönyökön you

A rövid

away

evening letter
to

such

on offers

He

it

other

my meg walls

way mastery
orders

a days er

to the

still

violence Project
uniformly

a Silence

Darinka been Alayna

While s

263

who

intentions she

the
mm commonly

taking

have you UNREST

married would rosszkedve

to

the and 7

nem God looked

exaggeration insight and

off

dotage must wilt

A

in now

oppose collection some

inches And

overheard We

January
place me

of father

mondani the two

It

egyel■re

effects
me later was

is beyond

accursed A

rows

child

last contiguity

She
engedj Some

Rousseau

would much themselves

what

a to treatment

so
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com