Harry Halpin: How did the idea of philosophical engineering come about?
Tim Berners-Lee: The phrase came about when we were originally discussing the idea of Web Science, and I was tickled by the fact that when you study and take exams in physics at Oxford, formally the subject is actually not physics but experimental philosophy. I thought that was quite an interesting way of thinking about physics, a kind of philosophy that one does by “dropping things and seeing if they continue to drop” – in other words, “thinking about the stuff you do by dropping things.” Then it came up again when trying to explain to people that when we design Web protocols, we actually get a chance to define and create the way a new world works. It struck me what we ended up calling “Web Science” could have been called “philosophical engineering,” because effectively when you create a protocol you get the right to “play God” and define what words mean. You can define a philosophy, to define a new world. So when people use your system, when they run the protocol, to a certain extent they have to leave their previous philosophy at the door and they have to join in and agree they will work with your system. So you can build systems, worlds, which have different properties. That's exciting, and a source of responsibility as well.
Harry Halpin: Would you consider the creation of Web standards to be an act of philosophy in progress?
Tim Berners-Lee: Certainly, when people write a specification, they argue about what words mean until everyone assumes that they mean in some sense the same thing. When the concepts in different people's brains have been sufficiently well-aligned and there have been enough connections between the concepts, this is written down in a language that people feel comfortable with and that they share. You can if you want to philosophically argue that a word is in fact ambiguous, but nobody bothers. Understand when you play the game [of specification-writing], you're not going to argue about that. For example, you're not going pay a bill online, and then afterwards come back and say “Well, I sent some HTTP headers off, but because their just HTTP headers, they don't actually mean anything.” As a spammer said once, “It's just a form field, I can put whatever I like there, it doesn't have to be the person sending the email.” But it does if your playing the game! I think one of the things we're missing is the relationship between the law of the land and protocols. It should be easier to establish that when someone disobeys a protocol that they've broken the law via a straight-forward path.
Harry Halpin: One of the most important aspects of natural language is that it's composed of words. In contrast, the Web is a space of URIs. How is it that URIs and their meaning differs from other possible systems like natural language? What is special about URIs?
Tim Berners-Lee: There are many URI schemes, but one thing that is nifty about HTTP URIs is that they have domain names in them. So they're hierarchical, and a domain is something that one can own. In the way the protocol works, the owner of the domain has the right to say, and the obligation to say on the Semantic Web, what the things in that domain mean. It's not a question of philosophical discussions between third parties. If there's a dispute about what a URI stands for, then the way the protocol works is that you go to the person who owns the domain name, who typically delegates it to someone else, who has in turn designed an ontology that they store on a Web server. The great thing about the Web is that you can look up the HTTP URI in real-time to get some machine-readable information about what it means straightaway.
Alexandre Monnin: Regarding names and URIs, a URI is not precisely a philosophical concept, it's an artifiact. So you can own a URI while you cannot own a philosophical name. The difference is entirely in this respect.
Tim Berners-Lee: For your definition of a philosophical name, you cannot own it. Maybe in your world, in your philosophy, you don't deal with names that are owned, but in the world we're talking about, names are owned. Some people have a philosophy where they they find it useful to think of a name as just a function of use, not of definition. Other people work in worlds like lawyers, where the model is that there is a definition, a legal definition of a term. There's enough law to insist that while meaning is use, but use is according to definition, because otherwise people could get put in jail. So there are models, and now we're adding another one, in which meaning is defined by the owner of a name.
Harry Halpin: Wasn't it controversial that when the Web was first starting that everything could be named with a URI?
Tim Berners-Lee: At the IETF certainly there was resistance. I originally called these things “Universal Document Identifiers” (UDIs) even before we started using them for concepts. The IETF were a bit put off, thinking it was too much hubris to call them “universal” but now I realize that I should have held firm and said “but they are,” as any alternative system of naming you can make out there, I can map it to the character set we use in URIs and I can invent a new scheme for it. So because we can map any scheme, we'd already mapped Gopher and FTP and these things. Now we've got HTTP and there will be lots of other schemes. So in a sense it is universal, we're saying anything, any name that you come across, can be mapped into this space. So yes, there was a lot of pushback against that, and hence the “uniform” rather than “universal” in URIs.
Alexandre Monnin: Given the origins of philosophical engineering and Web Science, don't you think that Web Science is doing two things? The Web is an artifact, we produce it, we implement it, and as you said, we decide what the protocol means and how it should be used. On the other other hand, Web Science is a science, so we make discoveries and we are also surprised by our own creation.
Tim Berners-Lee: The Web Science cycle starts off with idea that the design of the Web is not just the design of one thing but the design of two things; for example in email, there's a general technological protocol like SMTP and there's a social protocol. In email, there's a social protocol that states that everyone involved is ready to run a machine that has the space to store all your friends emails while they are en route to its destination, that people will send email to each other on perfectly reasonable topics, and that people will read email that they receive. There's a social piece of email, then e-mail is actually pushed around with SMTP and pulled off with IMAP, and those pieces then together form a system. It's a microscopic system that explains how one person sends another person an email through a finite number of hops, but then you get the effects of scale. So the engineering of Web Science is not like building a mousetrap. You design a microscopic system, but what you're interested in is the macroscopic phenomena that emerge. So when you do the science, the analysis, and the whole rest of the cycle for e-mail, you look at what is happening and notice “Spam has happened, oh dear!” What went wrong? One of our social assumptions was wrong, namely that everybody is friendly and will only send email to another person when the other person wants to read it. So the academic assumption is broken, and we have to redesign e-mail, and interestingly no-one has really succeeded in redesigning either the social or technical piece of mail to make spam go away. So there's an example: there's a design piece and an analysis piece, there's an engineering piece and a science piece, one being done on the microscopic system and the other being done on the microscopic system, and we're missing a lot of the mathematics that let us understand the connection between the two levels.
Harry Halpin: What is the role of philosophy in Web Science? Is there such a thing as a philoosophy of the Web?
Tim Berners-Lee: An awful lot of philosophy in the past has been in a wasted as it was done before we understood evolution. We were trying to understand emotions, and now we can point to evolution producing mammals with emotions. A lot of philosophy in the past is inapplicable. A lot people might say that philosophy is irrelevant to daily life, but if a Working Group stops and people start arguing about what things really mean, and people refuse to play the game, refuse to say what terms mean, and they don't do their job to define a protocol properly, then its a philosophical task to point out to them that this is important. Also, philosophy may be necessary to explain when the legal system hits the Web. When you make a web-page you can link to anything, you can write anything about it, but when a lawyer comes along and reserves the right to charge you to link to their page, then in a way it's a philosophical question, because you have to tie it to the way the protocol is defined over a name, just a reference, something that has never been controlled over the millennia. Systems where you control names haven't worked so far, and so you need the philosophy to show how these protocols ground out in legal history, in concepts for using names lawyers understand.
Alexandre Monnin: What do you expect from philosophy of the Web?
What I would like for philosophers to do is to work diligently and to produce very nice, simple, documents which describe to people like computer scientists how things work in a simple way. What happens when you click on a link? Quite a lot of that is philosophy. So, I'd for you to have enough of a body of understanding, so when people in a Working Group stop and say “Wait, this doesn't match what I learned from Wittgenstein” that you can say “No, please go read this pamphlet, its about philosophical engineering and it explains the philosophy of what your doing, so you won't find Wittgenstein very useful in this case or these are the bits that you will find useful.” So if you can produce enough discussion and understanding so that we don't have to stop work for philosophical discussions and we can rely on it being there, that would be excellent.
Editor’s note: This interview is to appear in "Philosophical Engineering", a collected set of writings from the PhiloWeb workshops published by Wiley-Blackwell press.