In two relatively well known books, A Treatise of Human Nature (1739) and An Enquiry Concerning Human Understanding (1748), David Hume explored the grounds upon which we make statements about unobserved phenomena. There are many examples in the schools of philosophy about Hume's speculations, the colour of Swans or the life sustaining qualities of bread. But whatever the example, everything boils down to one question: can we make a truth bearing observation about something that we have not observed?
A recent comment by the founder of Signal messenger is interesting because it offers an opportunity to put Hume's theory into practise and see if we can determine if the view is truth bearing:
People don’t want to run their own servers, and never will.
— Moxie Marlinspike
What struck me most about the statement has been the many, though not very varied, reactions. Those who agree seem to be letting it go largely un-commented. Those who disagree are seeking to cast it into the pot of rubbish predictions which have emanated from technology leaders over the generations by comparing it with other cherry picked examples. Notable in this genre of memorable quotes doing the rounds:
I think there is a world market for maybe five computers.
— Thomas Watson, president of IBM, 1943
There is no reason anyone would want a computer in their home.
— Ken Olsen, founder of Digital Equipment Corporation, 1977
I predict the Internet will soon go spectacularly supernova and in 1996 catastrophically collapse.
— Robert Metcalfe, founder of 3Com, 1995
While this trip down memory lane is enjoyable, it is also problematic. The reason is that this kind of criticism is falling foul of what is termed a category error.
The original example of a category error, or category-mistake, was given by Gilbert Ryle in his book The Concept of Mind (1949), in which a visitor to Oxford University is said to have asked, upon viewing the colleges and library: "But where is the university?" The error by the visitor was to assume that a university is part of the category 'units of physical infrastructure' rather than 'institutions'.
Taking the example quotes above about computers and the internet, the categories are 'projected use' — in the case of anticipated world market and why someone would want a computer in their home — and 'lifespan' — in the case of the internet going supernova and collapsing. Categories into which Moxie's assertion doesn't fit. But I will come to that in a moment.
For now, the reason the category in question is important to apprehend is because unless we do so, we can't accurately judge the veracity of a statement. A failure which may score politics points, but will likely result in a judgement based on faulty logic and personal bias rather than Hume's concept of inductive reasoning. A form of reasoning that allows us to speculate about something which hasn't been, or can't be, observed.
What People Want — By The Numbers
Stating what people want is to set oneself up for failure. To say that people want to stop climate change is only to surface people who deny climate change is real. To say that biological sex is a thing is only to surface people who deny a binary view of the biological world. In the case of this article, to say people don't want to run their own servers is only to surface the ~1% of people who do.
In what started as a well considered and important rebuttal of Moxie's position, André Staltz observed:
This [Moxie's statement] cannot be factually true, some people want to run their own servers. What is the thriving self-hosted open source community doing, other than running their own servers? Are the people running Mastodon instances, Yunohost apps etc not people?
While on firm ground in preferring to use 'percentages at scale' and acknowledging that blanket statements are always problematic because 'there is always someone interested in the weirdest and most niche topic', the article isn't the defence of a love of self-hosting it at first seems.
To give a personal example. I ran my own server for many months. In this context, I was a self-hosting statistic. But I didn't do this because I want to run my own server. Rather, I did it because I want to be admin of my own instance and run it using my own domain. This distinction between action and intent is important because to say 'N people run their own server' only demonstrates what people do. It does not account for a person's preference. A point clearly demonstrated in my own case when I switched to Masto.host.
I now have a way to run my own Mastodon instance without having to run my own server. An example of someone who shows up in the statistics of Mastodon instances, but who doesn't run their own server. A classic in the genre of a false positive in the percentages at scale data.
To return to the more fruitful ground of criticism; that Moxie's statement can't be factually true because 'people' must include all people and that it only takes one person who wants to run their own server to falsify the assertion. The critics of Moxie's belief should take heart as there are even more problematic elements in the statement 'People don’t want to run their own servers, and never will' than have been surfaced thus far.
Sui Generis
At the heart of this article is a statement about what hasn't been observed and, even more problematically, a statement that can't be observed. Namely, to check with everyone currently living and ask if they want to run their own servers and to ask everyone yet to be born if they want to run their own servers when the time comes. In other words, self-hosting statistics don't help as there could be entire groups of people who do want to run their own server, they just don't, much as someone in the future could decide self-hosting is the only way forward.
But to determine if Moxie's statement is 'true', we don't need to poll everyone alive or work out how to ask people yet to be born what are their future intentions. Instead, we simply need to understand the category of the statement and from that work out the balance of probability. To take a non-server example, do people want to cook their own food?
Certainly there are some people who love to cook and would do so, and in my experience do do so, even when they have a chef to prepare their perfect meal. Then there are people who, for security reasons, prefer to prepare their own food because they fear being poisoned. Then there are people who cook not because they want to, but because they need to eat and no one else is going to make their dinner. But in what category does cooking fall?
If it is the category of 'to prepare a meal', and if we assume that because they do prepare a meal then they want to prepare a meal, the statement 'no one wants to cook' is false. But if cooking is in the category of a 'medium', the process by which people receive a meal, then the case that no one wants to cook could be made. While this can always be statistically falsified in a specific instance, simply find someone who does want to cook their own food, it is still a statement which may be true as such because when given the option of having someone else prepare a beautifully cooked and perfectly safe meal, it isn't unreasonable to assert that most people would take the personal chef instead of slaving over a hot stove. And this is the heart of the conversation: not a statistical attempt to prove a zero adoption or interest rate, but to speak to the balance of probability when it comes to what people want. And for a tech leader, where future efforts should be made.
I would argue that servers in Moxie's statement fall into the category of an object, a physical box in a rack. Once correctly categorised, there is no inference in Moxie's statement that the world only needs a tiny number of servers, that there is no reason for an individual to want their own server, that servers will go boom and no one will use them, or that no one wants to control their own service. Nor does it mean that the statement carries with it the idea that people don't care about their security or privacy or that people aren't interested in content creation or running a service. Rather, it is to say people want a more sustainable way to manage their services than having to always be able to connect to the internet in case their box goes down; having to dial in at 2 am or by the side of the road because a service has gone offline; trying to figure out how to remotely manage a server from their phone in case of an outage.
For those who want to believe that self-hosting is the way of the future. In all likelihood there will always be the ~1% who will want to run their own server and will happily rummage through their draw for parts to come up with a MacGyver solution to an outage. But this doesn't falsify the statement 'People don’t want to run their own servers, and never will.' This is because the statement isn't about a statistical number and thus wrong so long as a small number of self-hosting enthusiasts remain. Rather, it is a political statement. That instead of investing more and more resources into solving for the problems of self-hosting, efforts should turn to solutions where the weakest link isn't 'do you have physical access to a box?'
In this context, self-hosting too often runs foul of elitism. The presumption that the only effective method of a safe, private and decentralised internet is for everyone to spin up a Raspberry Pi at home or turn their smartphone into a server. I do not knock, much less mock such efforts. In the face of them, I only recall the sage counsel of Thomas Sowell:
Freedom is not simply the right of intellectuals to circulate their merchandise. It is, above all, the right of ordinary people to find elbow room for themselves and a refuge from the rampaging presumptions of their "betters".
In this context, freedom is not the right of those of financial or computing privilege to circulate their self-hosting, but the right of ordinary people to find refuge from the rampaging presumptions of their technocratic 'betters' and enjoy privacy and security without having to run their own server.
Good night, and good luck.
Return to the Abbey by Eduardo Zamacois y Zabala (1841–1871) is licensed under Public Domain.
