Bill Gates' Shopping List to Build the Internet Desktop

by Jakob Nielsen on December 13, 1998

It is said that Windows'98 is so similar to Windows'95 because Apple hasn't invented anything worth copying since 1995.

It is time to kill off the Macintosh user interface and its clones. Rest In Peace: You Served Us Well. But we need a new user interface that is optimized for Internet access rather than office automation.

Apple's Sherlock search interface in Macintosh system 8.5 proves the benefits of integrating Web search with the operating system. But Apple is far from achieving the Internet Desktop: where's Cyberdog, where's OpenDoc, where's the ability to restrict searches to stuff your history list shows that you have seen before? For sure, calling something the iMac does not even come close to making it an Internet Macintosh . Instead, Apple has chosen a profit-maximization strategy to squeeze the last dollars out of its remaining loyal customer base by superficial innovations in hardware appearance.

Bill Gates has billions burning a hole in his pocket. Here's how I recommend he spend some of his excess money to acquire necessary elements for the Internet Desktop.

MilliCent: Micropayments

Micropayments are necessary for quality Web content. If I were given to conspiracy theories, I would speculate that Microsoft is keeping micropayments out of the operating system in order to cut off the air supply to the independent websites so that their own larger sites will rule the Web in the end.

A more likely explanation is that Microsoft was so focused on fighting Netscape that the IE team emphasized superficial page rendering over fundamental advances in Web quality.

Internet payment mechanisms have been a failure so far because of a lack of standardization. We all know that users don't want to download special plug-ins and configure strange special-purpose software. The only way to get micropayments widely used is to pick a single solution and integrate it fully with the basic Internet access software. Integration would not only make payments easier, it would also allow the page display mechanism (a software feature previously known as "a browser") to indicate the cost of links before users have clicked on them (for example by changing the cursor to one or more dollar signs, depending on the cost of following the link).

It probably doesn't matter which micropayment scheme is chosen, as long as Microsoft picks one and standardizes on it. But I have always been a fan of MilliCent from Digital. Now that Compaq has bought Digital, one can hope that MilliCent will be commercialized at a faster pace.

Google: Quality Search

Search is the most popular service on the Web because it fits the fundamental nature of the new medium: users choose where they want to go today.

Traditional search was based on finding all of the most relevant articles about the user's query. This approach worked well for scientists searching databases of research papers, but it has failed on the Web. We don't want all articles, and we don't even want the most "relevant" ones as determined by the number of times a certain keyword is used on a page. We want the best pages about a topic and the pages that are the best starting points for further hypertext navigation.

Google is a new search engine that explicitly includes the quality of pages when deciding what to include in a search results listing. Representing quality in the user interface is essential for the Internet Desktop and is the only way Web-wide searches can work.

Trellix: Authoring for the Online Age

Mainstream authoring software is all based on the concept of making the editing representation look like the final result, but WYSIWYG doesn't apply on the Web. The holy grail is no longer the production of long printed reports : instead the author's goal is the design of a hyperlinked navigation space.

Current Web authoring software is misguiding and leads authors astray with their focus on page appearance and single-page authoring. We need to move to true hypertext authoring as represented by the new Trellix tool.

Navigation-oriented authoring is particularly important for intranets where most content is contributed by people with no training in Web design. As long as intranets are dominated by print documents, corporate productivity will suffer huge hits (I estimate $50 billion per year).

Tcl and/or Frontier: Web Scripting

Internet programming languages have received a lot of attention, but most of the interactive aspects of the Web are better implemented in scripting languages than in full-featured programming languages.

Two scripting solutions that seem particularly promising are Tcl (by John Ousterhout who is a former Sun Microsystems Distinguished Engineer and one of the world's top software engineering geniuses) and Frontier (by Dave Winer who is a leading authority on good use of the Web and author of much revolutionary software). I am not a good enough programmer myself to choose or to architect the scripting solution for the Web, but I do know that we need one. The spirit of the Web requires fluent and dynamic evolution of interactive behaviors that cannot be locked down in compiled software.

What You Can Do While Waiting for BillG

Only Bill Gates can decide to throw away WinOff and build the Internet Desktop. Considering how much money he makes from Office, this may not happen any time soon. The Justice Department would do users a great favor if they stopped worrying about the integration of browsing into the operating system (a development that was shown to be beneficial by hypertext research in the 1980s) and started worrying about the Windows-Office ( WinOff ) platform.

Luckily the vast majority of the modern user experience is defined by Web content, with software playing a much smaller role. Thus, even though users definitely would be better off with Internet-optimized software instead of Office-optimized software, we can get part of the way there ourselves while waiting for Gates.

  • Plan your long-term Internet strategy on making money from micropayments: it will happen sooner or later and anybody who relies on advertising revenues will die as click-through rates hit 0.1% in 2001
  • Embed quality ratings in your own design (a primitive approach is the way I highlight the most-visited pages in my list of Alertbox columns; a better idea is to doctor site-wide search engines to put the best pages on top, regardless of keyword count)
  • Teach your staff to write for the Web and to construct navigation spaces instead of long linear reports (nobody has time to read a long report anyway)
  • Install a good scripting solution on your Web server

Share this article: Twitter | LinkedIn | Google+ | Email