muflax65ngodyewp.onion/content_blog/hack/xmonad.mkd

---
title: Meditation on XMonad
alt_titles: [XMonad]
date: 2010-05-03
techne: :done
episteme: :discredited
---

Ignorance is the root of all suffering - ignorance about reality,
about what is. By holding wrong assumptions, we create false
expectations and false needs. [^0] 

I will not reflect on large parts of reality, but only a small one: window
managers (WM). [^1] 

The most basic ignorance about WMs is the ignorance about
their existence. The computer does not just show data to us, but it can show it
to us in any way we want. It is this basic understanding that leads to the first
conclusion: If the way data is shown to us is lacking, it is not our fault, but
the computer is not doing it's job properly. Furthermore, if we have to spend a
lot of time just telling the computer how we would like to see something, we are
actually doing someone (or rather, something) else's work.

Therefore, tiling WMs. If you arrange your own windows, why are you using a WM
at all? Wouldn't it be more honest, instead of saying "I'm running Windows / KDE
/ OS X to show my windows", to admit "Windows / KDE / OS X is running me to show
it's windows."? Sure, the computer can not read your mind and some occasional
hints might be necessary, but the less work you do, the better. 

Desire creates suffering. This is maybe the most misunderstood of Buddhist
truths. People hear "desire creates suffering" and think "What? Is this going to
be a moral how material possessions are bad for me?  Money, cars, houses and so
on lead to greed, obsession, and so on.  I get it.". This is not what this is
about. The problem with desire is not the desire itself. It is not a problem
that we want to be happy, to be rich and so on. The problem is, instead, that
what we want is impossible. Our desires *fail* us. We are mistaken about the
nature of reality and expect the wrong things. We think that money could make us
happier, so we want more of it, but it ultimately won't. From wrong assumptions
you can only get bad results. 

In retrospect, I can see this clearly now on my journey to a better window
manager. It was my unwillingness to let go of old habits, my wrong ideas about
what I really need or want, that made adopting a new WM hard. So I went first to
WMs that offered great configuralibity and many features. "You can do anything
you want!" But this lead to useless features and distractions. It is only now
that most WMs fail me because my hardware setup is a bit tricky, that I can
understand. Only now when I understand better what my brain really needs, do I
grow tired of those full of bad design.  Xmonad, in a way, is peace for me. It
is mathematical in nature.  The fact that it is written in Haskell might seem
like a gimmick at first, but the connection is in fact very deep. I understand
now that it could not have been written in anything else. Xmonad exemplifies the
idea of purely functional programming. "Normal" programming is almost always
imperative - the programmer tells the computer how to do something. But in
functional programming, the hacker instead tells the computer what something
*is*. This is a profound difference.[^3] 

In any other customizable WM I have ever used, I would create a complex
configuration to tell it exactly what I wanted it to do in some case or another.
I would do the bulk of the lifting, so to speak, either by constantly adjusting
the windows the WM handled wrong or by writing elaborate procedures to automate
this work. But with Xmonad, this is different. It is not my job to figure out
how to arrange windows, so I should never have to tell my WM anything about
this. The only thing I ever have to tell it is about what is. I should never
write something like "to go to the next tag, you read in all tags, sort them,
filter some out, find the current one and then shift to the next in the list".
I instead write: "I want the next non-empty, non-visible tag now".  I give
Xmonad a few simple hints and that is it. "If it's name is in this list, I want
it floating. If I'm currently out of space here, try a different screen. There
is a status bar I'm running, so be careful not to overlap it." 

For the first time, I feel like my WM is actually intelligent and wants to help
me. It is not my slave, not my servant who follows my orders. It does not look
down on me, thinking itself smarter than me, only an obstacle to its flawless
performance. Instead, Xmonad is my friend. It understands window handling and
can take care of it. I only tell it some personal preferences. If it doesn't
think I need something, it is probably right. 

It is astonishing how easily we pick up delusions.  We see something once and
think it should always be that way.  Rarely do we question "Is this really
necessary? Is there no other way?" 

For me, those are some of the delusions that clouded my
judgement about WMs. 

"I need space! I want to see my desktop wallpaper!" What for? Have I not
something better to do than to stare at pretty pictures? 

"I want to tell my WM what window is in the foreground and what in the
background." The very concept is wrong. There is no "foreground" with focus -
you either see something or you don't. A window you can not read might as well
not be there at all. 

"I understand now, I use a tiling WM. But I want to control what window is
where!" Why? The very idea of a tiling WM is that the WM figures out what to
show you and how. You simply tell it what application has your focus right now
and what other applications belong to it (by giving them all the same tag /
workspace). 

"Xmonad has no stacked layout like wmii! I can not easily put dozens of windows
in one column!" Why would you do this in the first place? You certainly can not
read them all. Let Xmonad only show you the ones that matter and search for
other ones if you need them. Or think about grouping them better. Why open 20
PDFs in separate windows if your viewer can take care of that? 

"Xmonad has no title bars.[^4] I will miss those!" Are you sure? What do you use
them for? The window content itself tells you what the window is.  If the
content is not visible, then a title bar will only waste space. If you need to
find something, let the WM do it for you. If you want status reports, use
notifications. 

By embracing not complexity, but simplicity, confusion ends. The best solution
to a problem is to make it obsolete - as Gordon Bells said, "The cheapest,
fastest, and most reliable components are those that aren't there.". 

By concentrating not on *how*, but on *what*, false
desires disappear. By letting go off false desires, suffering ends.

<%= image("guru.png", "Guru Meditation") %>

[^0]: Yeah, I have been reading Buddhist philosophy and history
    again. Can you tell? 
    
[^1]: The old monks have understood one thing: Truths about reality must be
    visible everywhere. There can not be any aspect of reality that is not
    permeated by them. Thus, we can improve our efforts by just focusing on one
    simple object.  Traditionally, one's breath, a candle or a rock have served
    this purpose. Some Zen traditions use 只管打坐 (shikantaza, "simply correct
    sitting") for this. If you can't understand reality just by sitting down and
    concentrating, then reality can't be understood at all.  Therefore we must
    be able to see all those Buddhist observations in everything we use,
    including the most fundamental GUI software - our window manager. 

[^3]: The classical example to demonstrate this is Quicksort. If you have ever
    programmed something, Quicksort was probably among it, but just to help you
    remember, I'm gonna tell you again what Quicksort is. We define Quicksort
    recursively like so: An empty sort is always sorted. To sort a list with at
    least one element, we take the first element (called the pivot) in the list
    and then separate the rest into two lists, one containing all the elements
    that are smaller and one containing all that are larger than the pivot. Now,
    to get the sorted result, we simply sort the first list, than add the pivot
    and finally add the sorted second list. Think about how you would solve this
    in an imperative language. In C, it would go something like this:

    ~~~
    #!c
    void swap(int *a, int *b)
    {
      int t=*a; *a=*b; *b=t;
    }
    void sort(int arr[], int beg, int end)
    {
      if (end > beg + 1) {
        int piv = arr[beg], l = beg + 1, r = end;
        while (l < r) {
          if (arr[l] <= piv)
            l++;
          else
            swap(&arr[l], &arr[--r]);
        }
        swap(&arr[--l], &arr[beg]);
        sort(arr, beg, l);
        sort(arr, r, end);
      }
    }
    ~~~

    This is a typical example - we tell the computer exactly what to do to get
    the result we are interested in. But remember I said that in a functional
    language, we tell the computer what something *is*. I already told you what
    Quicksort is, so let's write this down in Haskell:
    
    ~~~ 
    #!haskell
    qsort [] = []
    qsort (x:xs) = qsort lesser ++ [x] ++ qsort greater
         where lesser  = [y | y <- xs, y < x]
               greater = [y | y <- xs, y >= x]
    ~~~ 
    

    And that's it. 

[^4]: Technically, you can add them, but they are not normally there.