What does the history of the web tell us about its future?

Following my invitation to speak at the WWW@20 celebrations [my bit starts about 133 minutes into the video] – this is my attempt to squash the most interesting bits into a somewhat coherent 15 minute presentation.

20 years ago Tim Berners-Lee was working, as a computer scientist, at CERN. What he noticed was that, much like the rest of the world, sharing information between research groups was incredibly difficult. Everyone had their own document management solution, running on their own flavour of hardware over different protocols.  His solution to the problem was a lightweight method of linking up existing (and new) stuff over IP – a hypertext solution – which he dubbed the World Wide Web – and documented in a memo “Information Management: A Proposal“.

Then for a year or so nothing happened. Nothing happened for a number of reasons, including the fact that IP, and the ARPANET before that, was popular in America but less so in Europe. Indeed senior managers at CERN had recently sent out a memo to all department heads reminding them that IP wasn’t a supported protocol – people were being told not to use it!

Also because CERN was full of engineers everyone thought they could build their own solution, do better than what was already there – no one wanted to play together. And of course because CERN was there to do particle physics not information management.

Then TimBL got his hands on a NeXT Cube – officially he was evaluating the machine not building a web server – but, with the support of his manager, that’s what he did — build the first web server and client. There then ensued a period of negotiation to get the idea out freely, for everyone to use, which happened in 1993. This coincided, more or less, with the University of Minnesota’s decision to charge a license fee for Gopher. Then the web took off especially in the US where IP was already popular.

first web server
The Worlds first Webserver.

The beauty of TimBL’s proposal was it’s simplicity – it was designed to work on any platform and importantly with the existing technology. The team knew that to make it work it had to be as easy as possible. He only wanted people to do one thing, that one thing was to give their resources identifiers – links – URIs; so information could be linked and discovered.

This is then is the key invention – the URL.

To make this work URLs were designed to work with existing protocols, in particular it needed to work with FTP and Gopher. That’s why there’s a colon in the URL — so that URLs can be given for stuff that’s already available via other protocols. As an aside, TimBL’s said his biggest mistake was the inclusion of // in the URL — the idea was that one slash meant the resource is on the local machine and two somewhere else on the web, but because everyone used http://foo.bar it means the second / is redundant. I love that this is TimBL’s biggest mistake.

He also implemented a quick tactical solution to get things up and running and demonstrate what he was talking about — HTML. HTML was originally just one of a number of supported doctypes – it wasn’t intended to be the doctype but HTML took off because it was easy. Apparently the plan was to implement a mark-up language that worked a bit like the NeXT application builder. But they didn’t get round to it before Mosaic came along with the first browser (TimBL’s first client was a browser-editor) and then it was all too late. And we’ve been left with something so ugly I doubt even it’s parents love it.

The curious thing, however, is that if you read the original memo — despite its simplicity — it’s clear that we’re still implementing it, we’re still working on the the original spec. Its just that we’ve tended to forget what it said or decided to get sidetracked for a while with some other stuff. So forget about Web 2.0.

For example, the original Web was read-write. Not only that but it used style sheets and a WYSIWYG editing interface — no tags, no mark-up. They didn’t think anyone would want to edit the raw mark-up.

The first web site was read and write
The first web site was read and write

You can also see that the URL’s hidden, you get to it via a property dialog.

This is because the whole point of the web is that it provides a level of abstraction, allowing you to forget about the infrastructure, the servers and the routing. You only needed to worry about the document. For those who remember the film War Games — you will remember that they had to ‘phone up individual computers — they needed this networking information to access the computer, they needed to know its location before they could use it. The beauty of the Web and the URL is that the location shouldn’t matter to the end user.

URIs are there to provide persistent identifiers across the web — they’re not a function of ownership, branding, look and feel, platform or anything else for that matter.

The original team described CERN’s IT ecosystem as a zoo because there were so many different flavours of hardware, different operating systems and protocols in use. The purpose of the web was to be ubiquitous, to work on any machine, open to everyone. It was designed to work no matter what machine or operating system you’re running. This is, of course, achieved by having one identifier, one HTTP URI and defererence that to the appropriate document based on the capacities of that machine.

We should be adopting the same approach today when it comes to delivery to mobile, IPTV, connected devices etc. — we should have one URI for a resource and allow the client to request the document it needs. As Tim intended. The technology is there to do this — we just don’t using it very often.

The original memo also talked about linking people, documents, things and concepts, and data. But we are only now getting around to building it. Through technologies such as OpenID and FOAF we can give people identifiers on the web and describe their social graph, the relationships between those people. And through RDF we can publish information so that machines can process it, describing the nature of and the relationship between the different nodes of data.

Information Management: A Proposal
Information Management: A Proposal by Tim Berners-Lee

The original memo described, and the original server supported, link typing so that you could describe not only real word things but also the nature of the relationship between those things. Like RDF and HTML 5 now does, 20 years later. This focus on data is all a good idea because it lets you treat the web like a giant database. Making computers human literate by linking up bits of data so that the tools, devices and apps connected to the web can do more of the work for you, making it easier to find the things that interest you.

The semantic web project – and TimBL’s original memo – is all about helping people access data in a standard fashion so that we can add another level of abstraction – letting people focus on the things that matter to them. This is what, I believe, we should be striving for for the web’s future because I agree with Dan Brickley, to understand the future of the web you first need to understand it’s origins.

Don’t think about HTML documents – think about the things and concepts that matter to people and give each it’s own identifier, it’s own URI and then put in place the technology to dereference that URI to the document appropriate to the device. Whether that be a desktop PC, a mobile device, an IPTV or third party app.