My involvement with the early days of the Web

Dave Raggett W3C@10, Boston 1st December 2004

I got involved in work on hypertext in 1990 when I was a researcher at HP Labs in Bristol, England. We were interested in exploiting graphical hypertext for user interfaces to expert systems for ordering HP computer systems. Our inspiration came from Bill Atkinson's Hypercard. The project (called the Protek Pilot, after the name of an HP distributor based near Heathrow airport), was very successful and whetted my appetite for business applications of hypertext. It featured wizards for guiding product selection, rich forms, and generated quotes ready for faxing to customers.

A shift in corporate priorities led to the abandonment of the project at the point where we were about to start on making it work across the network. I didn't despair and decided to see if I could find other people with a similar interest in global hypertext. I soon came across Tim's name and arranged to visit him in CERN in the Fall of 1992.

We got on really well talking about ideas for enriching HTML and HTTP to the point where they could be used for applications like HP's Protek Pilot. When I got back to HP Labs, I started to develop a browser and an HTTP server to test out the ideas, and joined in the discussions on the www-talk mailing list that Tim had set up in CERN.

By July 1993, this had led to an Internet Draft for HTML+. Later that year I was invited by Peter Flynn to Cork, Ireland for a WWW/TEI meeting [1] with browser developers, including Chris Wilson from the NCSA Mosaic team. We discussed ideas for tables and forms. The pace continued to pick up. In 1994 I took part in three conferences relating to the Web, an HP internal conference in Palo Alto, the first international World Wide Web conference in CERN, and the Internet Society conference in Prague, as well as starting an IETF working group to formalize HTTP.

I was invited by Yuri Rubinsky to the SGML Open's Summer Camp at the Xerox Document University in Virginia. Yuri was very much aware of the disruptive nature of the Web and wanted to build a bridge between the SGML and Web developer communities. The CALS table model[2] was too complex for the Web browser developers and a simpler approach proposed in the HTML+ discussions on won out. SGML Open later became OASIS. Yuri has since died, but is much missed. My kids remember him for his magic tricks at a restaurant meeting in Toronto.

After setting up W3C in October 1994, Tim made plans to move to the USA. I had been lobbying within HP for an assignment to work with Tim, and visited him at CERN a few days before he was due to leave for America, with his house full of packing crates, Tim introduced me to Francois Fluckiger who was taking over from Tim. The plan was for me to be posted to CERN to help with the European activities of W3C.

A few months later, Robert Cailliau and Francois Fluckiger organized a meeting in Brussels at the European Commission, to discuss proposals for a European Web consortium. I met Jean Francois Abramatic for the first time, and Al Vezza stole the show when he appeared hot footed from the first USA meeting of the MIT Web Consortium. The people present at the Brussels meeting were strongly in favor of a single international effort rather than separate regional ones.

A few days later CERN backed out of the running, and Tim started looking for another European host. Tim asked me to attend a meeting at Oxford University as one of the possibiities. Both Tim and I had been students at Oxford at the same time. We hadn't met there, although we had a mutual friend in Paul Rouse. Paul and I were working on our doctorates at the Department of Astrophysics. Tim told me that INRIA were in the front running, and asked me what I thought about working at Rocquencourt in the south of Paris.

I instead found myself being posted by HP to MIT in the Spring of 1995. As well as chairing the new IETF HTTP working group, I was busy working on HTML 3.0 having written the draft specification while on a family vacation in the English Lake district. This followed from an agreement with Tim Berners-Lee and Dan Connolly at CERN just before the 1st International World Wide Web conference. Dan would focus on formalizing existing practice as HTML 2.0, while I would focus on the next generation and new features such as tables forms, and embedded objects. HTML+ would be repositioned as HTML 3.0.

The IETF HTML working group decided that HTML 3.0 was too large a specification to work on and that it would be better to work on specifications for individual features. I therefore switched my attention to focus on HTML tables. After lots of work, we were able to advance it as an IETF experimental RFC. As this was going on, the browser wars got going, with Microsoft entering the running and giving Netscape some real competition. Concerns were raised about the risk of the Web splintering as browser developers continued to innovate with new tags, so that web pages designed for Netscape's browser wouldn't work on Microsoft's and vice versa.

At MIT we decided to take action, and with Susan Westhaver's help I arranged for a meeting at Spyglass's offices in Schaumberg Illinois. We invited representatives from IBM, Microsoft, Netscape, Sun Microsystems, and Spyglass. We chatted about requirements for embedded objects with each company having a different perspective. The meeting proved valuable and we formalized the arrangement for future meetings using the brand new Editorial Review Board process set up for the earlier W3C work on PICS (Platform for Internet Content Selection).

The HTML ERB started by looking at new features drawing for inspiration on the ideas in HTML+ and HTML 3.0. We then realized that a higher priority was to provide an updated version of HTML 2.0, which by now looked pretty dated. We weren't sure what to call this new version, and as 3.0 had been taken, we ended up with HTML 3.2, as no one wanted the stigma of a dot one version. HTML 3.2 was very successul as it provided a safe set of tags that would run in a predictable fashion on both Netscape Navigator and Internet Explorer.

With HTML 3.2 out of the way we started work on the next major release of HTML, factoring in our earlier work on the object tag, as well as richer forms and tables drawing upon work in the IETF HTML working group. The W3C Process itself continued to evolve and the HTML ERB was re-launched as the HTML Working Group, focusing on the development of HTML 4.0.

This left open the relationship between the IETF and the W3C. It was apparent that the W3C was proving a more effective forum for work on HTML. On the advice of John Klenzin, IETF Applications Area Director, the IETF HTML working group closed down in 1996, and I handed over the role of chair for the IETF HTTP working group to Larry Masinter from Xerox PARC. It took several years for browsers to implement the features of HTML 4.0, but it has stood the test of time. Along the way the W3C HTML working group has spun off new groups for work on scripting (the Document Object Model), CSS for styling web pages, MathML for mathematical expressions, and more recently for work on XForms, voice browsers and multimodal interaction.