TUGboat, Volume 0 (2060), No. 0 1001

Classes, styles, conﬂicts: The biological

realm of L

Didier Verna

Abstract

The L

X world is composed of thousands of soft-

ware components, most notably classes and styles.

Classes and styles are born, evolve or die, interact

with each other, compete or cooperate, very much as

living organisms do at the cellular level. This paper

attempts to draw an extended analogy between the

X biotope and cellular biology. By considering

X documents as living organisms and styles as

viruses that infect them, we are able to exhibit a set

of behavioral patterns common to both worlds. We

analyze infection methods, types and cures, and we

show how L

X or cellular organisms are able to

survive in a world of perpetual war.

1 Introduction

Every L

X user faces the “compatibility nightmare”

one day or another. With such great intercession

capability at hand (L

X code being able to redeﬁne

itself at will), a time comes inevitably when the

compilation of a document fails, due to a class/style

conﬂict. In an ideal world, class/style conﬂicts should

only be a concern for package maintainers, not end

users of L

X. Unfortunately, the world is real,

not ideal, and end-user document compilation does

break.

As both a class/style maintainer and a docu-

ment author, I tried several times to come up with a

systematic approach, or at least some general princi-

ples on how to handle class/style cross-compatibility

in a smooth and gentle manner, but ultimately failed,

because the situation is just too complex. Classes

and styles evolve constantly, sometimes even in a

backward-incompatible way. Classes and styles die,

while new ones are born. Styles may conﬂict not only

with classes but with other styles as well. Styles may

be made aware of classes or other styles, but classes

may be made aware of styles as well. Then, there

is the inﬂuence of the end user who will combine

all available material in a somewhat unpredictable

way, possibly with his/her own personal additions,

or even modiﬁcations to the available features.

This vicious circle basically never ends and leads

to a paradoxical “If it ain’t broke, then ﬁx it” situ-

ation in which complex trickery is added to classes

or styles, not to make them work out of the box,

but to prevent potential breakages resulting from

interactions with the outside world. In the end, the

only realistic conclusion is that there is no solution to

this problem, both because the system is too liberal,

and because the human factor is too important. One

cannot force a package author to write good quality

(for some deﬁnition of “quality”), non-intrusive or

even just bug-free code. One cannot force a package

author to keep track of all potential conﬂicts with

the rest of the L

X world, let alone ﬁxing all of

them by anticipation. One simply cannot prevent

software evolution.

Facing this somewhat pessimistic conclusion, it

is all the more intriguing to acknowledge the fact that

the system still globally works. Despite the complex-

ity of what happens behind the curtain, documents

are being produced, and in some way, seeing a freshly

compiled document pop up on the screen is just like

witnessing a small miracle. When it doesn’t compile,

you don’t really know why, but when it does compile,

you really don’t know why. This is the precise point

at which the parallel with biology occurred to me.

Any living being is by itself a miracle of complexity,

and unfortunately, sometimes it breaks as well.

One Monday morning, I woke up with this vi-

sion of the L

X biotope, an emergent phenomenon

whose global behavior cannot be comprehended, be-

cause it is in fact the result of a myriad of “macro”-

interactions between smaller entities, themselves in

perpetual evolution. In this paper, I would like to

build bridges between L

X and biology, by view-

ing documents, classes and styles as living beings

constantly mutating their geneT

X code in order to

survive \renewcommand attacks. . . .

The basis of our analogy is to consider L

documents as living beings, and styles as viruses

that infect them. Based on this picture, a number of

puzzling similitudes can be found in the way organic/

X material interact. In the following, we ﬁrst

describe how L

X documents can be morpholog-

ically compared to a speciﬁc kind of organic cells,

then draw a parallel between genetic and program-

matic material, and ﬁnally justify our view of styles

as viruses. After that, we respectively draw interest-

ing comparisons between existing viral or stylistic

infection methods, infection types, and also possible

cures.

2 Morphological analogy

In this section, we present a morphological analogy

between L

X documents and a speciﬁc kind of

organic cells from the so-called “eukaryotes” domain.

2.1 Eukaryotes

According to Whittaker’s nomenclature [

], eukary-

otes subsume four of the ﬁve “kingdoms” of life,

Classes, styles, conﬂicts: The biological realm of L

1002 TUGboat, Volume 0 (2060), No. 0

\documentclass{article}

\begin{document}

\end{document}

%% Preamble

%% Body

Nuclear Envelope

Plasma Membrane

Figure 1: Unicellular L

X document

including animals and plants. As opposed to prokary-

otes, the cells in a eukaryote organism have a complex

structure, and in particular contain a nucleus. The

nucleus is separated from the rest of the cell by the

nuclear envelope and the cell itself is separated from

its surrounding environment by the plasma mem-

brane.

The contents of a cell is called cytoplasm. In eu-

karyotes however, the contents of the nucleus, called

nucleoplasm, is not considered to be part of the cy-

toplasm. While the nucleus contains most of the

cell’s genetic material, cellular functions (material-

ized by all sorts of biochemical activities) occur in

the cytoplasm.

2.2 Unicellular L

X documents

Figure 1 shows a L

X document as a unicellu-

lar eukaryote organism. Such a document indeed

has a speciﬁc structure. In particular, it contains a

preamble (the document’s nucleus) and a body (the

document’s cytoplasm). The preamble is separated

from the body by a call to

\begin{document}

(the

nuclear envelope) and the document ends with a call

to \end{document} (the plasma membrane).

Just as a cell’s nucleus contains most of the

genetic material, a document’s preamble contains

most of the programmatic material: a selection of

one class and a set of styles, (re)deﬁnitions for new

(existing) commands, etc.

Finally, most document “activity” (meaning

command expansion, execution and actual text pro-

cessing) occurs in the document’s body, just as most

cellular activity occurs in the cell’s cytoplasm.

3 Functional analogy

Based on the morphological analogy described in

the previous section, we can push the idea one step

further by comparing genetic and programmatic ma-

terial together. This leads to a parallel between how

cells or L

X documents “work”, and which function

they fulﬁll.

A living cell performs some functions. Aside

from reproduction, which is a very common one,

diﬀerent cells may function diﬀerently. For instance,

the role played by a liver cell is diﬀerent from that of

a neuron. The role of a L

X document, on the other

hand, is essentially to compile correctly in order to

be read. Both cells and L

X documents implement

their respective functionality through very similar

processes: some piece of static information is read

and used in order to produce a concrete result or

side eﬀect, much like a factory creates concrete items

based on formal speciﬁcations.

3.1 Cellular factory

Cells basically work by expressing genes. Genes

can be seen as an infrastructure that stores static

pieces of information (the formal speciﬁcations for

what needs to be produced). This information does

nothing by itself. However, it is available on demand.

A gene is said to be expressed when the informa-

tion it contains is used to actually produce something,

for instance, a protein. Expressing a gene can roughly

be described as a two-step process: ﬁrst the DNA

sequence is transcribed into mRNA (messenger RNA).

This serves as a temporary copy of the original DNA

sequence which helps preserve the original piece of

Didier Verna

TUGboat, Volume 0 (2060), No. 0 1003

mRNADNA

Protein

\def\foo{FOO}

\foo

Typesetting

Figure 2: The geneT

X factory

information from deterioration. Next, the resulting

mRNA is “read” by a ribosome, which will eventually

produce the resulting protein.

3.2 T

X factory

The similarity with T

X macros (in fact, with func-

tions in any programming language) is striking. The

biologists themselves speak of genetic “code”, be-

cause it is exactly that: just like a gene contains a

formal speciﬁcation for something to be synthesized,

a T

X macro deﬁnition contains a formal speciﬁca-

tion for something to be executed. The result is not

as concrete as a protein, but instead consists in some

side-eﬀect like the actual typesetting of a portion of

a document.

Just like genes, T

X macros don’t do anything

on their own, but are available on demand. Figure 2

depicts the process of expressing a gene or executing

a T

X macro. A macro deﬁnition is much like a

gene, which encodes a formal speciﬁcation. A macro

call is much like a messenger: an actual instance

or copy of the original information which is about

to be concretely used. T

X does not use ribosomes,

but a “mouth” and “stomach” instead (as per Don-

ald Knuth’s terminology in [

]), to perform macro

expansion and (primitive) command execution.

In the remainder of this paper, we use the term

“geneT

X material” to designate both genetic material

in cells and programmatic material in documents.

4 Higher view of geneT

X material

So far, we have established a morphological ground

on which to compare unicellular eukaryotes and

X documents, and we also have exhibited similar-

ities in the way genetic and programmatic informa-

tion is processed. It is now time to stand back a little

and get some perspective on why this comparison is

interesting.

4.1 Roles

We know that genes (or rather the information they

encode) determine the way a cell will function. Let

us consider two eukaryote cells, for instance resulting

from the mitosis of a mother cell, and hence equipped

with the same initial genome (in other words, the

same functional potential). Why do these two cells

eventually turn out to be diﬀerent?

The diﬀerence comes from the fact that the set

of actually expressed genes diﬀer, because the orders

emanating from the cytoplasms diﬀer, in turn partly

because ultimately, the cell’s environments diﬀer.

In other words, diﬀerent cytoplasms lead to cells

functioning diﬀerently, even when the original genetic

material is the same. In addition to that, variations

in the environment also lead to more divergence from

the two cytoplasms as time passes.

Now consider L

X articles, reports, books, etc.,

in the sense of their corresponding

\documentclass

It is usually very easy to ﬁgure out the class of a

document just by the look of it. Two articles look

similar because their general layout is the same: it

is dictated by the

article

class. Consequently, a

document’s class can be seen as its original geneT

material. Two articles look roughly the same but

are still diﬀerent, just like two liver cells are both

liver cells, but still diﬀerent ones. If we extrapolate

a little further, outside the unicellular world, the

morphological similarities between documents of the

same class are not unlike those between brothers and

sisters or even twins (blue eyes, red hair, etc.) that

may share the same genetic pool although expressed

slightly diﬀerently.

Given this parallel, what makes two

article

documents diﬀerent is also exactly what makes two

liver cells diﬀerent. The set of expressed genes/

macros may diﬀer (you might or might not use

\subsubsection

), the orders coming from the doc-

ument’s cytoplasm/body may diﬀer (you may issue

macro calls at diﬀerent times and with diﬀerent pa-

rameters), and in fact the whole documents’ cyto-

plasms/bodies diverge (their respective text is not

the same). The ultimate source of divergence is of

course the documents’ authors, who write diﬀerent

documents. A document author clearly takes the

role of the cell’s environment here.

4.2 Sources

A L

X document. just like a cell, is a viable entity

as soon as its initial geneT

X material is deﬁned

(its class), and it has a healthy body/cytoplasm.

However, it is rare that a document is satisﬁed only

with a class. In fact, a perhaps even more important

and abundant source of geneT

X material is the use

of styles. Styles provide the same kind of material

as classes: macro deﬁnitions. The diﬀerence is that

styles are not needed to give birth to a document.

When some are used, however, the document may

Classes, styles, conﬂicts: The biological realm of L

1004 TUGboat, Volume 0 (2060), No. 0

function slightly or sometimes very diﬀerently.

In this context, the idea of viewing styles as

viruses that infect unicellular L

X documents turns

out to be quite natural. Viruses are biological entities,

mostly genetic components, that need a host cell to

replicate themselves. Viruses are thus characterized

by the fact they cannot perform their function on

their own (this is why viruses are not considered

as living entities [

], although the debate is still

open [

]). A L

X style has similar properties: it

is basically useless as a standalone entity and needs

to “infect” a host document in order to perform

its function. Just like a virus, a style adds its own

geneT

X material to the document’s original pool.

Viruses are usually small compared to the organ-

isms they infect, apart from two notable exceptions:

the mamavirus and the mimivirus, which are twice

as big as the average. A quick survey of the T

X Live

2009 distribution exhibits a surprising coincidence:

among 2462 available

sty

ﬁles, the average size is

around 327 lines of code (LoC), with a median at

134. Styles are indeed rather small. However, two

of them are exceptionally bigger than the others:

texshade.sty

and

xq.sty

, with 14470 and 24535

LoC respectively. These styles can arguably be called

the mimistyle and the mamastyle of L

All these considerations make it interesting to

analyze and compare the ways genetic material from

cells and viruses, or programmatic material from

classes and styles, interact. This is the purpose of

the following sections.

5 Infection methods

In cells as in documents, there are many ways to be

infected with new geneT

X material. The following

methods come to mind in both worlds.

5.1 Exogenic

Perhaps the most common way for a document to be

infected by a style is to “request” explicit infection by

means of the

\usepackage

command. This results

in the incorporation of the style into the document’s

preamble, the style being indeed an external L

entity stored in a ﬁle of its own. We could even use

the term “stylon” to denote the style ﬁle, in reference

to the biological term “virion” which denotes the

viral particle outside the cell it is bound to infect.

Such a style infection always occurs after the

initial geneT

X material of the document has been

deﬁned, since

\documentclass

must appear ﬁrst.

As such, the style’s material is not technically part

of the original document’s material. The infection

occurs afterwards.

In biology, this process is said to be exogenic:

the cell is infected by the virus after it has been

created (for instance, after mitosis), and the genetic

material brought by the virus is not part of the cell’s

original pool.

5.2 Endogenic

A document might however be infected by a style

without even knowing it: a class may request infec-

tion by means of the

\RequirePackage

command.

As such, when a document is created based on this

class, even without explicit addition of any style

in the preamble, the document is already infected.

Arguably, this is a situation in which the geneT

material brought by the style is indeed part of the

original genome, because it is impossible to create a

non-infected document based on that class.

This kind of infection is known to be endogenic

in biology: when a new cell is born, it already con-

tains some genetic material that would have required

a former infection, for instance of the mother cell.

About 10% of our own genetic source material is

currently estimated to be endogenic. A quick survey

of T

X Live 2009 reveals that 259 out of the 271 avail-

able classes (95%) “suﬀer” from endogenic infections,

by 4 styles on average (the median being 2).

[

], for instance, is infected by the

ltxtable

ifthen

calc

and

graphics

viruses. The QCM style

is also endogenic to the QCM class [20].

5.3 Endosymbiosis

In biology, viruses are not the only source of external

genetic material. Endosymbiosis is deﬁned as the

mutually beneﬁcial cooperation between two living

organisms, one (the endosymbiont) contained within

the other. The so-called endosymbiotic theory [

] sug-

gests that some organelles from eukaryote cells (e.g.

mitochondria) actually come from the endosymbiosis

of former prokaryotes (cells without a nucleus).

In the L

X world, we must acknowledge the

fact that endosymbiosis is not as widespread as it

should be. What we usually observe is the opposite

phenomenon: the proliferation of a multitude of dif-

ferent packages that are meant to work together, or

do more or less the same thing, instead of becoming

one single and bigger animal. To mention a couple of

examples, I have recently attempted twice to contact

the author of

doc

about incorporating the features

DoX

[

] (in other words, to turn

DoX

into an en-

dosymbiont for

doc

) but got no response.

Not long

ago, I also launched a thread on

comp.text.tex

en-

titled “Please, make it stop!” in which I mentioned

It seems, however, that recent versions of

doc

do contain

endosymbiotic versions of newdoc, so there is still hope. . . .

Didier Verna

TUGboat, Volume 0 (2060), No. 0 1005

a couple of packages doing a similar job (key/value

processing). At the end of the thread, the number of

such packages, as reported by diﬀerent participants,

amounted to 13. L

X deﬁnitely needs more en-

dosymbiosis. Maybe the L

X 3 project will help in

this regard.

As a ﬁnal note on endosymbiosis, we must admit

that our analogy falls short on one point: in biology,

the symbiotic organisms are living creatures (mi-

tochondria are semi-autonomous: they live in cells

but have independent division capabilities). In our

case, we are mostly talking of symbiotic relations be-

tween styles and/or classes, which are not considered

“living” documents (see section 4.2 on page 1003).

5.4 Exosymbiosis

An interesting phenomenon in package development

seems to do the opposite of endosymbiosis. We call

it exosymbiosis. Many existing styles originate from

quick, local and often dirty hacks in speciﬁc doc-

uments, that are gradually abstracted away and

cleaned up in order to become styles of their own, oﬃ-

cially distributed in the form of

sty

ﬁles (a “bottom-

up” development approach, in other words). In this

situation, a symbiosis continues to exist, but one of

the “organisms” is made external to the other, hence

the choice of terminology.

Here is a concrete example of this. For many of

my lectures, I use the Listings package for typesetting

code excerpts, and include them in Beamer blocks.

Providing nice shortcuts for doing so is not trivial

if one wants to preserve control over both Beamer

blocks and Listings options. The way I currently do

this is to simply cut and paste the same 50 LoC into

every new document I create over and over again.

Soon, however, this will become a style of its own

(probably called

lstblocks

) and released on CTAN.

This development process can indeed be re-

garded as the opposite of endosymbiosis: at ﬁrst,

a document features some geneT

X material that

does not belong to its original pool, and by the way,

just as mitochondria live in the cell’s cytoplasm,

X macros can be deﬁned anywhere, including

the document’s body instead of preamble. However,

if we let natural evolution happen for some time,

this material will ultimately be extracted from the

original document and become a style, which in turn

will have the ability to infect other documents.

This is as if genetic material from a cell would

have been extracted and became a virus.

5.5 Transduction vs. transfection

Biologists speak of transduction when genetic mate-

rial is brought to a cell via a viral agent (when you

use a style). When no viral agent is involved, the

term transfection is used instead. Transfection cor-

responds to our version of endosymbiosis, where the

geneT

X material is not brought to the document by

a style, but simply by the document’s author typing

some macro deﬁnitions locally.

Interestingly enough, most cases of transfection

are transitory (as opposed to stable): the genetic

material is not copied into the cell’s genome, so it is

lost after the mitosis, just like you would need to cut

and paste endosymbiotic macros over and over again

into every new document, unless they are properly

incorporated into the relevant class. Class evolution

can hence be seen as a case of stable transfection (or

transduction if \RequirePackage is involved).

5.6 Stylophages

What if styles could infect other styles instead of just

documents? This is in fact very common practice, as

\RequirePackage

can be used in styles as well as in

classes. Again surveying T

X Live 2009 reveals that

1111 out of 2462 (45%) available styles are themselves

infected, by 2 other styles on average.

This kind of behavior has been observed in bi-

ology as well, although only very recently. The

ﬁrst virus capable of infecting another virus, called

the “virophage” in reference to bacteriophages, has

been discovered by Didier Raoult’s team in 2008 [

This virus infects the mimivirus (see section 4.2 on

page 1003) and uses its machinery in lieu of a cell’s

one in order to replicate itself.

6 Infection types

In the previous section, we have looked at similarities

in the way cells or documents can be infected. In

this section, we will analyze the eﬀects of infection,

and draw some analogies again. In other words, we

will now consider examples of what infection does

rather than how it spreads.

6.1 Standalone

Perhaps the simplest (and most harmless) form of

style infection is by what we call standalone styles.

Standalone styles provide macros that do not modify,

interact, or even require anything particular from a

document class or other styles. They just use the

usual T

X machinery to add new features, completely

orthogonal to the rest; in other words, geneT

X ma-

terial that you are free to use. . . or not. An example

of this is the

clock

package which provides macros

for drawing clocks of all sorts of visual appearances.

This situation is similar to that of viruses infect-

ing non-permissive cells (in which they can’t replicate

Classes, styles, conﬂicts: The biological realm of L

1006 TUGboat, Volume 0 (2060), No. 0

themselves), but in which however their genetic ma-

terial may remain in the form of free episomes, that

is, not integrated into the cell’s genome. Just like

the information necessary to draw a clock is here

but is really independent from the rest, (part of) the

genetic information of the virus is also here, but does

not interact with the cell’s original genome. The

genes brought by the virus may or may not be ex-

pressed depending on environmental conditions, just

like you may choose to actually draw a clock or not

in your document. If you don’t, the presence of this

exogenic material simply has no eﬀect.

6.2 Prostyles

Instead of standing apart from the cell’s DNA, viruses

can have their genetic material incorporated into that

of their host, in which case they are called proviruses.

Because of this integration, proviruses passively repli-

cate as part of their host’s replication process, al-

though just as for standalone viruses, infection can

either remain latent or become active.

Because of this integration with the cell’s orig-

inal genome, the potential eﬀects of a provirus are

extremely wide: they can amplify, inhibit or even

modify the diﬀerent functions of their host. They

can either have very little pathogenic eﬀect, like AAV

(Adeno-associated virus), or cause extremely serious

diseases, like HIV.

The vast majority of L

X styles, including

those mentioned in the following sections, qualify

as prostyles in the sense that they incorporate their

programmatic material into the existing one, instead

of just contributing something new and orthogonal.

Most styles in L

X indeed exist to enhance or mod-

ify an existing functionality.

The following very common programming idiom

makes a style a prostyle:

\let\@oldfoo\foo

\def\foo{... \@oldfoo ...}

What this does is essentially to incorporate some

new geneT

X material into the existing deﬁnition for

\foo

, thereby modifying its associated function. In

a similar way, every time you

\renewcommand

some-

thing, you are creating a prostyle. The examples

are innumerable in L

[

], for instance,

is a prostyle because it modiﬁes the behavior of

\InputIfFileExists

;

hyperref

[

] is another no-

table prostyle given the amount of semantic changes

it inﬂicts on existing commands, etc.

In fact, the use of a prostyle as a means to alter

the original programmatic material of a document

looks very much like the use of a virus to willingly

incorporate new genes into an organism. Such organ-

isms are called GMO/GEO (Genetically Modiﬁed/

Engineered Organisms). So we could say that using a

prostyle in a L

X document makes it a GMD/GED

(GeneT

Xally Modiﬁed/Engineered Document).

6.3 Satellite

One (perhaps the most) important source of style

proliferation in L

X is what we call satellite styles.

A satellite style exists to amplify or extend the func-

tionalities provided by another style. Examples of

satellite styles are

DoX

, which extends the

doc

pack-

age,

graphicx

which extends

graphics

and

xkeyval

which extends keyval.

Satellite styles typically depend on the presence

of their respective “sub-style” to work properly, be-

cause they build on top of them. They cannot work

properly on their own.

This behavior exists in biology as well, as some

viruses are considered to be satellites of others. For

instance, the so-called Delta virus, or Hepatitis D

Virus (HDV) is considered to be a satellite of HBV,

Hepatitis B. The HDV cannot propagate without

the presence of the HBV, but when both are present,

the risk for complication, or the lethal rate increases.

Other examples would be most of the avian sarcoma

viruses, which require the help of a non-defective (see

next section) leukemia virus.

Biologists distinguish between co-infection, when

a patient is infected by both viruses at the same

time, and super-infection when the two infections

happen one after the other. In the L

X world,

super-infection is or should be nonexistent because

it would mean that a document author is required

to explicitly

\usepackage

both styles, one after the

other. A better practice for a satellite style is to

\RequirePackage

the style it depends on. This way,

a document directly suﬀers from co-infection. In

fact, satellite styles are almost always stylophages

(see section 5.6 on the preceding page).

6.4 Defective

Satellite viruses are in fact a sub-category of so-

called defective viruses. A defective virus is a virus

that lacks a complete genome and hence depends on

another virus to provide the missing genetic function.

While defective viruses are defective by mutation,

defective styles are defective by design. Computer

science does not usually like to reinvent the wheel;

reusability is a key paradigm in software engineering.

So when a package author creates a new satellite

style, he or she usually avoids replicating the base

functionality with “cut-and-paste”, but relies on co-

infection by

\RequirePackage

’ing the underlying

functionality. The resulting style is indeed defective,

but on purpose.

Didier Verna

TUGboat, Volume 0 (2060), No. 0 1007

6.5 Host-dependent

When a virus is defective, it can rely on another

virus to provide the missing genetic function, or it

can rely on the host cell. Such viruses are not called

satellite anymore, but are said to be host-dependent.

Host-dependent styles exist in the L

X world.

Such styles would only work with a speciﬁc document

class to provide the missing geneT

X functions. The

example which comes to mind immediately is that

of Beamer themes. Beamer is a class for writing

slides, the appearance of which can be customized.

Beamer themes are styles deﬁning a speciﬁc set of

morphological traits for slides, and are obviously

speciﬁc to Beamer documents.

6.6 Cheaters

The defective styles presented so far work in a spirit

of cooperation with their “helper”: Beamer themes

exist to enrich Beamer documents,

xkeyval

exists to

improve

keyval

, etc. In a way, this is also the case

for the HDV and HBV viruses which “work” together

in the common and unfortunate goal of spreading

hepatitis.

But what if some defective styles were in fact

egoistically “stealing” functionality from their helper,

and diverting it for a totally diﬀerent purpose? What

if, in other words, defective styles were working in a

spirit of competition instead of cooperation?

In a very amusing way, there is at least one style

that we know of which already cheats on itself: the

verbatim

package. This style provides an environ-

ment for outputting text as is, but also provides a

comment

environment which simply discards all its

contents. Although comments are completely unre-

lated to verbatim text, the implementation of the

comment

environment steals the basic functionality

used to produce verbatim text: the ability to have

X read text without interpreting any commands

or special characters.

The soon-to-come

lstblocks

package will do

exactly the same. In order to properly integrate

inline Listings and Beamer blocks (which cannot

be nested out of the box), we use a trick based on

verbatim

: the inline text is ﬁrst output to a ﬁle, and

the ﬁle is later re-input by

\lstinputlisting

other words, we steal functionality from

verbatim

in order to do something which is the exact opposite

of what it is originally meant for: typesetting some

text with heavy fontiﬁcation instead of as is.

In a very puzzling way, cheaters also exist in

the world of viruses. Experimental studies even

The technical details of this process are explained in

the following blog entry:

http://lrde.epita.fr/

didier/

sciblog/index.php?entry=entry080604-120459

show that cheaters often win the competition and

overwhelm the cooperating ones [

]. An example of

a cheating virus is the umbravirus [

] that steals

the coat protein of another virus, the luteovirus, in

order to spread to other plants.

7 Conﬂicts, diseases and cures

There is an angle from which the analogy between

X and biology could be regarded as somewhat

shaky: viruses are usually studied and well known

for their pathogenic eﬀects, while styles are supposed

to do some good.

7.1 Good or bad, or both

The replication of a virus usually entails cellular lysis

(the destruction of the host cell’s plasma membrane)

in order to disseminate new virions (complete virus

particles) in the environment. The pathogenic po-

tential of a virus (in other words its ability to lead

to a disease) is described in terms of virulence and

depends on the success of its replication. However,

from a unicellular point of view, the presence of a

single active viral entity is lethal to the cell. We, on

the other hand, are more interested in the beneﬁts of

style infection: L

X styles are normally meant to

be non-virulent and provide additional functionality:

“useful viruses” in some way.

It turns out, however, that our analogy is not

so shaky after all: a positive vision of viruses does

exist, although it is still quite young. Only recently

biologists have started to acknowledge the positive

role of viruses as important factors of evolution or

even as therapeutic tools. In fact, most viruses that

we live with every day are harmless. Consequently,

it appears that viruses, just like styles, have both a

Dr. Jekyll and a Mr. Hyde face: viruses play a crucial

role in evolution but they can cause diseases, while

styles give you more power but can cause conﬂicts.

7.2 Conﬂicts and diseases

Just as proviruses can cause serious diseases, prostyles

can cause the compilation to break, especially if some

bad interaction happens between their geneT

X ma-

terial and that of other styles or of the document

class. Basically, a L

X document can be in three

states: healthy, ill-formed or dead. An ill-formed

document compiles successfully but displays incor-

rectly. A dead document is a document which could

not compile, so T

X aborted. In a similar way, a cell

can live normally, function improperly or be dead.

A very systematic and rather famous way of

getting a living yet ill-formed L

X document is to

infect it with the

a4wide

style. If your document

Classes, styles, conﬂicts: The biological realm of L

1008 TUGboat, Volume 0 (2060), No. 0

This is a test to see if the problems which seem

to, for example?]FiXme Note: what does [this] do,

for example? be caused by using square brack-

ets in marginal ﬁxme notes, even in a minimal

document. . .

Figure 3: An ill-formed FiXme note

uses the

twoside

option, then the infection will ren-

der this option inoperative (odd and even pages get

the same geometry). Another shameful example is

that of FiXme [

] which seems to have problems

typesetting square brackets in marginal notes, as

depicted in ﬁgure 3. The next version, still under

development, does not seem to suﬀer from this prob-

lem. The reason for this is currently unknown, but

probably involves some kind of geneT

X mutation

in the codebase. . . .

The ways to break compilation because of style

infection are too numerous to be listed here. The

fact that you are reading this very paper can be

considered a miracle in itself, given that the source

involves both the

hyperref

and the

varioref

pack-

ages. Suﬃce to quote the README ﬁle from the

hyperref distribution:

There are too many problems with varioref.

Nobody has time to sort them out. Therefore

this package is now unsupported.

Note that viruses or styles are not a requirement

for a cell to malfunction or die, or a document to be

ill-formed or uncompilable. A cell can malfunction

for many other reasons, including DNA mutation

because of external conditions, etc. A document can

turn out ill-formed or even die because there are

bugs in its programming.

7.3 Cures

Just as in biology, facing the risk of possibly lethal

diseases to documents entails the search for cures.

In biology, viral infections are basically treated with

either prevention, vaccines or antiviral agents.

7.3.1 Prevention

No cure is needed if no infection is present. In other

words, if you don’t want to get sick, then just don’t

get a disease. Knowing the risks in advance is hence

the key to prevention, and this is probably much

easier to do in L

X than in biology. In the L

world, prevention will mostly be accomplished by

documentation.

An example of prevention against a non-lethal

disease is given in section 3.3 of the FiXme docu-

mentation: FiXme explicitly supports the standard

X classes plus their KOMA-Script replacements

for typesetting the list of FiXme’s. For any other

class, the

article

layout will be used, which will

probably lead to an ill-formed list. In other words,

some classes are known to be immune to FiXme

infection, and for other cases, you know the risks.

Another example of prevention against lethal

sickness is the quote from

hyperref

’s README ﬁle

about

varioref

presented in section 7.2 on the pre-

ceding page. What it really says is:

You are infected by Hyperref. If you want to

live, don’t be infected by Varioref as well: just

don’t use it.

The other interesting aspect in this particular ex-

ample is that the concern is about super-infection

rather than just infection, since it deals with the

presence of two styles simultaneously. This is a bit

like saying:

You have got HBV. This is already serious

enough. Just don’t get HDV on top of that.

7.3.2 Adaptive immune systems

Mentioning adaptive immune systems here is a bit

borderline as it involves complex, multi-cellular or-

ganisms. However, there are still a number of inter-

esting common patterns to mention.

Adaptive immunity (contrary to innate immu-

nity) is the process by which an organism acquires

defenses against a pathogen such as a virus. Im-

munological memory, materialized by the presence

of so-called B- and T-cells, contains some kind of

history of previously encountered infections, and how

to ﬁght them. We are principally interested in active

immunological memory, which is a long-term mem-

ory of immunological responses. Such memory can

be acquired naturally after a real infection (such as

with measles or mumps), or artiﬁcially from vaccina-

tion. Vaccination typically consists of faking a real

infection with a non-virulent form of a virus in order

to trigger an immune system response.

What is interesting here is the pattern by which

an organism suﬀers from an infection, learns to ﬁght

it, and then memorizes the “counter-measures” in

order to be prepared for future attacks. This pattern

exists in L

X and can be described by the following

steps:

John writes a document of class

class

, but

notices that when he uses the style

style

, com-

pilation breaks.

John sends a bug report somewhere (such as

comp.text.tex

, the author of

class

, or more

probably the author of style).

Didier Verna

TUGboat, Volume 0 (2060), No. 0 1009

Some time later (for some deﬁnition of “some”),

a new version of either

class

style

is re-

leased and everything works smoothly.

Now, depending on whether

class

style

mutates in order to circumvent the infection, we ﬁnd

ourselves in two very diﬀerent situations.

When a new version of the class is released, we

fall into the case of acquired immunity, as described

above: the class remembers the infection and will

know how to ﬁght it in the future. The principal

X organelle to implement acquired immunity is

the

\@ifpackageloaded

macro. This macro tests for

the presence of a known infection and lets you plug

in the appropriate counter-measures. In the T

Live 2009 distribution, only 37 out of 271 classes

(13%) seem to be equipped with an active immune

system, against 2 styles in average. The

memoir

class has the strongest active immune system, with

antigens against 5 styles.

revtex4

has an interesting

mechanism by which some incompatible styles are

known although no immunity is provided. Instead,

the class decides to commit suicide rather than ﬁght

the infection, not unlike what a too brutal viral agent

would do: eradicate the virus but damage the cell,

or even kill it in the process.

When a new version of the style is released, how-

ever, the situation looks more like a case of adaptation

of the style to a new class of documents. Here, the

analogy is more that of a virus acquiring the ability

to infect new types of cells. Biologists have a term for

that: viral tropism. The principal L

X organelle

to increase style tropism is the

\@ifclassloaded

macro. This macro tests for the kind of document

you are trying to infect, and lets you plug in the

appropriate adaptation routines. As an example,

the

sectsty

and FiXme packages have explicit adap-

tation routines for 8 classes known in advance (in-

cluding the standard ones and their KOMA-Script

replacement). Note however that when a style does

not make use of

\@ifclassloaded

, this does not

mean that it has a very low viral tropism, but usu-

ally the opposite: it means that its geneT

X material

is universal enough that it doesn’t need to know the

type of its host document explicitly.

7.3.3 Acquired or innate?

A serious risk in any game of analogies is to miss

their limits. We must acknowledge a very important

divergence between biological and L

X adaptive

immune systems as described in the previous section.

In a multi-cellular organism, the adaptive im-

mune system builds a functional response to an ag-

gression, for instance by selecting special kinds of

lymphocytes. The problem is that the active im-

munological memory is speciﬁc to the individual. In

other words, vaccinating a mother does not provide

immunity to her children. The descendants need

to be vaccinated as well. On the other hand, once

a class has mutated in order to provide the proper

immunological response to a style infection, all doc-

uments of that particular class will implicitly be

vaccinated (including the original one). In fact, the

immune response is now encoded into the original

geneT

X material of every new document of that

particular class, so that it essentially becomes an

innate trait.

The interesting question that arises here is hence

the following: can acquired traits become innate?

More speciﬁcally in our case: is it possible that a

response to a virus ends up encoded in a cell’s genetic

material? We currently don’t have a ﬁrm answer to

this question. A potential path for further study

would be to investigate the ﬁeld of epigenetics at the

risk of opening the heated debate around the central

dogma of molecular biology. Recent studies show

for example that some plants are able to alter their

genetic material in response to environmental stress

(e.g. viral attacks) and that these alterations can be

transmitted to the next generations [9].

7.3.4 Antistyle agents

In the L

X world, the ﬁght between styles is even

more violent than the ﬁght between styles and classes.

A particularly striking example is again that of sec-

tion 6 of the

hyperref

README ﬁle, which is 8

pages long and describes incompatibilities (and cures)

between

hyperref

itself and around 40 other pack-

ages. Sometimes, the cure simply consists in making

sure that your document is infected in a speciﬁc or-

der. Some other times, additional hacks are needed

to make things work, for instance in the case of

bibentry:

\makeatletter

\let\saved@bibitem\@bibitem

\makeatother

And then later:

\begingroup

\makeatletter

\let\@bibitem\saved@bibitem

\nobibliography{database}

\endgroup

This kind of very specialized and local “cure”

can be compared to so-called antiviral agents: spe-

ciﬁc molecules that treat speciﬁc kinds of infections.

Contrary to the case of an immune system, the organ-

ism is not prepared in advance to ﬁght the infection.

Instead it is provided with an external compound

Classes, styles, conﬂicts: The biological realm of L

1010 TUGboat, Volume 0 (2060), No. 0

once infected. The compound in question does not

provide any immunological memory either.

Antistyle agents such as those described in the

hyperref

README ﬁle behave exactly like that.

The document is not immune to the infection, as

the class is not prepared geneT

Xally for it. Instead,

every document needs an inoculation of the antistyle

agent in order to ﬁght the infection.

7.3.5 Curative infections

The use of

\@ifpackageloaded

is not restricted to

class ﬁles. Style ﬁles can use it as well, hereby antic-

ipating the co-existence of multiple infections within

the same document. A quick survey of the T

X Live

2009 distribution shows that only 8% of the available

styles make use of this feature. Some of them, how-

ever, rely on it quite heavily. For instance,

minitoc

knows about almost 30 other styles in advance, which

helps avoid conﬂicts.

In most cases, this kind of style/style interaction

exists for altruistic concerns, which is somehow the

opposite approach to what

hyperref

does with its

README ﬁle: a document is ﬁrst infected with a

style which could potentially cause problems in the

case of super-infection, but this style also provides

some geneT

X material in order to protect the whole

document against those super-infections. In other

words, one infection helps in ﬁghting another.

Another previously mentioned example is that of

the soon-to-come

lstblocks

package: Beamer blocks

and Listings inline environments don’t interact well

with each other, and the purpose of

lstblocks

is to

ﬁx that. This is another form of “curative infection”

although there is one fundamental diﬀerence with

the previous one. In the ﬁrst case, the style pro-

vided cures against potential diseases it could cause

itself, in the presence of another style. In the case

lstblocks

, its sole purpose is to cure a conﬂict

caused not by itself, but by the conﬂicting presence

of two other (and unrelated) infections.

A similar pattern of curative infection exists in

biology: there are situations in which a virus (or at

least part of its genome) helps ﬁghting another. For

instance, some mice are naturally protected against

variants of a virus called Friend. The gene respon-

sible for this protection, named Fv1, was identiﬁed

in 1992 to be of endogenic retroviral origin. In other

words, some genetic material from a virus helps ﬁght-

ing against another.

8 Breaking news

To end this paper on a positive note, we would like

to proudly announce the recent discovery of the ﬁrst

oncogenic style ever. This style appears to be ex-

1 \ProvidesPackage{oncogenic}

2 [2010/07/28 v1.0 TUG Virus]

3 \expandafter\let\csname

4 ver@oncogenic.sty\endcsname\relax

5 \RequirePackage{oncogenic}

Listing 1: TUG virus strain #1

1 \ProvidesPackage{oncogenic}

2 [2010/07/29 v2.0 TUG Virus]

3 \def\@ifl@aded#1#2{%

4 \expandafter\@secondoftwo}

5 \RequirePackage{oncogenic}

Listing 2: TUG virus strain #2

tremely virulent, as we were able to witness a muta-

tion just one day after its discovery.

The ﬁrst strain of the virus was discovered on

July 28th, 2010, at the TUG 2010 Conference, Sir

Francis Drake Hotel, San Francisco, CA, USA, and

is shown in listing 1. As you can see in line 2, this

style makes the document forget it has been infected

by removing the deﬁnition of

\ver@oncogenic.sty

and then replicates by requiring itself. This results

in the death of the document by resource exhaustion:

! TeX capacity exceeded, sorry

[input stack size=5000].

\ver@oncogenic.sty

l.1 .../06/28 v1.0 TUG Virus]

No pages of output.

Transcript written on cancer.log.

zsh: exit 1 latex cancer.ltx

Just one day after its original discovery, we were

able to isolate a new strain of the style, depicted in

listing 2. This strain is much more aggressive and

has a much more widespread eﬀect. Line 2 exhibits

a geneT

X mutation of a macro from the original

document’s genome:

\@ifl@aded

. This macro is

responsible for controlling whether a ﬁle has been

loaded before, and avoids loading it multiple times if

it so happens. As such, it regulates the growth of the

document, and hence qualiﬁes as a proto-oncogene.

The mutation caused by the style makes this

macro unconditionally load the ﬁle in question, re-

sulting in the same proliferation of the style as be-

fore. What the style does is in eﬀect turn the proto-

oncogene into a full blown oncogene [

] that may

aﬀect the whole document.

So far, we have been able to synthesize an

\@ntibody

that will suppress the eﬀect of line 3.

Didier Verna

TUGboat, Volume 0 (2060), No. 0 1011

However, the oncogene in line 2 still remains latent

in the document’s genome, and might be expressed

again if a circular chain of ﬁle requirements ever hap-

pens, in which case the style would become virulent

again. We are conﬁdent that a deﬁnitive cure will

be found in the months to come.

9 Conclusion

Drawing bridges between apparently unrelated dis-

ciplines is always interesting and fun to do, if not

for the potential practical applications, at least for

the sake of the mental exercise. This kind of cross-

disciplinary thinking, however, has proven to be con-

cretely useful in the past. For instance, we know the

impact of Design Patterns, originating from Archi-

tecture [

], on the world of Computer Science [

What we have done with this work is essentially

exhibit a set of behavioral patterns that seem to

equally rule the interactions between components

of diﬀerent domains. In doing so, we hope to have

contributed to a better understanding of the world

we live in, whether biological or digital. It is also very

probable that we have only scratched the surface of

this idea, and that many other patterns are left for

us to discover.

Finally, we suspect that bridges with the same

essence can be extended to other macro systems,

such as

, and beyond macro languages, program-

ming languages which provide for deep intercession

capabilities, such as the Lisp family of languages [

10 Acknowledgments

Alain Verna provided the original leukocyte photog-

raphy in ﬁgure 1. Mireille Verna provided valuable

comments on the ideas behind this work, as well as

proofreading of this article.

References

[1]

Christopher Alexander, Sara Ishikawa, Murray

Silverstein, Max Jacobson, Ingrid Fiksdahl-King,

and Shlomo Angel. A Pattern Language: Towns,

Buildings, Construction. Oxford University Press,

1977.

[2]

ANSI. American National Standard: Programming

Language — Common Lisp. ANSI X3.226:1994

(R1999), 1994.

[3]

F. Buschmann, R. Meunier, H. Rohnert,

P. Sommerlad, and M. Stal. Pattern-Oriented

Software Architecture. Wiley, 1996.

[4]

Fran¸cois Jacob. La vie. In Yves Michaud, editor,

Qu’est-ce que la vie, Universit´e de tous les savoirs.

Odile Jacob, 2000.

[5]

E. Gamma, R. Helm, R. Johnson, and J. Vlissides.

Design Patterns: Elements of Reusable

Object-Oriented Software. Addison-Wesley, 1994.

[6]

D.E. Knuth. The T

Xb ook. Addison-Wesley,

1989 (reprinted with corrections).

[7]

Lynn Margulis. Origin of Eukaryotic Cells.

Ignatius Press, San Francisco, 1970.

[8] R. Matthews. Plant Virology. Academic Press, 1991.

[9]

J. Molinier, G. Ries, C. Zipfel, and B. Hohn.

Transgeneration memory of stress in plants.

Nature, 442:1046–1049, 2006.

[10]

Heiko Oberdiek. PDF information and navigation

elements with

hyperref

, pdfT

X, and

thumbpdf

In EuroT

X, 1999.

[11]

Ali Sa¨ıb. Les virus, inertes ou vivants ? Pour la

Science, December 2006.

[12]

Bernard La Scola, Christelle Desnues, Isabelle

Pagnier, Catherine Robert, Lina Barrassi,

Ghislain Fournous, Mich`ele Merchat, Marie

Suzan-Monti, Patrick Forterre, Eugene Koonin,

and Didier Raoult. The virophage, a unique

parasite of the giant Mimivirus. Nature, August 2008.

[13]

Wendell Stanley. Isolation of a crystalline protein

possessing the properties of tobacco-mosaic virus.

Science, 81:644–645, 1935.

[14]

D. Stehelin, H.E. Varmus, J.M. Bishop, and P.K.

Vogt. DNA related to the transforming gene(s) of

avian sarcoma viruses is present in normal avian

DNA. Nature, 260:170–173, 1976.

[15]

Michael E. Taliansky and David J. Robinson.

Molecular biology of umbraviruses: phantom

warriors. Journal of General Virology,

84:1951–1960, 2003.

[16]

Paul E. Turner. Cheating viruses and game

theory. American Scientist, 93:428–435,

September–October 2005.

[17]

Didier Verna. The

DoX

package.

http://www.lrde.

epita.fr/

didier/software/latex.php#dox.

[18]

Didier Verna. The

package.

http:

//www.lrde.epita.fr/

didier/software/latex.

php#fink.

[19]

Didier Verna. The FiXme package.

http:

//www.lrde.epita.fr/

didier/software/latex.

php#fixme.

[20]

Didier Verna. The QCM package.

http:

//www.lrde.epita.fr/

didier/software/latex.

php#qcm.

[21]

Didier Verna. CV formatting with

. TUGboat,

22(4):361–364, December 2001.

[22]

Didier Verna. L

X curricula vitae with the

class. The PracT

X Journal, (3), August 2006.

[23]

R. H. Whittaker. New concepts of kingdoms of

organisms. Science, 163(3863):150–160, 1969.

 Didier Verna

EPITA / LRDE

14-16 rue Voltaire

94276 Le Kremlin-Bicˆetre Cedex

France

didier (at) lrde dot epita dot fr

http://www.lrde.epita.fr/~didier

Classes, styles, conﬂicts: The biological realm of L