No Comments Written by Anders on March 28, 2008 – 10:46 am

The intent was to make more accessible the various documents being generated at CERN – the European research agency in Geneva that runs the particle accelerators for the physicists studying sub-atomic structure. Like any other large organization, CERN generates numerous reports – management reports, planning reports on proposed future uses of the accelerators, experiment reports, and endless tables of results from experiments. Berners-Lee had been a long-term proponent of hypertext systems, and his vision was to have all these reports linked into a hypertext web, with readers able to navigate via links in the reports. In late 1990 he was given the chance to develop a demonstration system, working with the aid of a vacation student.

The success of Berners-Lee’s approach is largely due to its simplicity. There had been earlier attempts at creating hypertext systems, but these had always resulted in complex, proprietary systems. Berners-Lee chose to use plain text documents, annotated with a simple markup language – the language that eventually became the Hypertext Markup
Language (HTML). Markup languages of various forms had then been in use for some 15 years; a common use was to annotate plain text by inserting markup tags that conveyed formatting information. Many different word-processing and text-display programs can handle documents in this form. There was an agreed standard for markup languages (SGML – the Standard Generalized Markup Language) that specified how markup tags should be defined and used. Berners-Lee simplified things (a little too much in some places – HTML violates some of the SGML rules for a good markup language). He defined the basic tags that are now familiar to all (even to primary schoolchildren who these days compose web pages). His documents had head and body sections delimited by appropriate tags; formatting tags that allowed section headers of various forms to be specified; and limited paragraph controls, lists, tables, and display controls that change fonts and so forth. The really smart part was to use markup tags for the hypertext links to related documents. The formats have changed a little, but these ‘anchor’ tags are familiar to web page authors:

<a href=source>Next month’s scheduled accelerator downtimes</a>

The program that displayed the hypertext (the browser in modern terminology) used some mechanism to highlight the tag and to allow a user to activate this link. Such an action replaced the current document in the browser with the linked document. The source documents for the other reports could have been on any of the machines in the CERN empire. Consequently, the source references in the hypertext links couldn’t simply be file names for local files. They had to identify the source machine as well as the filename. When a link was activated, the browser program would have to connect to a file
server program running on the identified machine.

The ftp file transfer protocol has been around since about 1973 (well before TCP/IP and the Internet; it has just evolved as underlying technologies have changed). The ftp protocol allows a client to login, establish a connected session, and then transfer multiple files; it actually makes use of separate control and data transfer connections using two TCP/IP links between client and server. Berners-Lee felt that ftp was too heavyweight a solution. His hypertext users would be reading a report, would get to a link and activate it, and receive a new report that they would read for many minutes before again connecting via another link (possibly a link to a different machine). There was no need for ftp’s login sessions, state maintenance and capabilities for transfer of multiple files.

I would like to know other opinions on serving static hypertext, as well as finding out more about this.