In 2005, the PRASA (Pattern Recognition Association of South Africa) conference website was a single index.html file with hyperlinks to each paper’s abstract page. To add a new paper, a graduate student had to open the file in Notepad and insert a new <li> element — a task that many in the lab jokingly called “programming HTML.” This phrase, technically incorrect but widely used, captures how researchers in the mid-2000s treated HTML: as a simple, hands-on tool for disseminating their work before content management systems became common.
HTML as a Publication Medium for Pattern Recognition Research
Throughout the mid-2000s, HTML served as the primary vehicle for sharing pattern recognition research online. Conference proceedings, lab member directories, dataset documentation, and even interactive demos were built with hand-coded HTML pages. The PRASA symposia, for instance, often published their programs and abstracts as a set of static HTML files hosted on university servers. Researchers who wanted to contribute had to follow a strict markup template — a process that required attention to closing tags and relative paths rather than any algorithmic logic.
This hands-on approach meant that many graduate students and faculty members became proficient in HTML, even if they never wrote a line of JavaScript or PHP. The act of creating a <table> to list accepted papers or a <form> to collect feedback was seen as a practical skill, often taught informally in lab meetings. For under-resourced language research groups, where funding for dedicated web developers was scarce, HTML was the lifeline that kept their findings visible to the international community.
The Misnomer of “HTML Programming”
Strictly speaking, HTML is a markup language, not a programming language. It defines structure and presentation, not logic or computation. Yet the phrase “program HTML” persisted in academic circles, reflecting a broader confusion that also surrounded early web technologies. In pattern recognition labs, where researchers were accustomed to writing MATLAB scripts and C++ code, the shift to a declarative syntax felt like a different kind of programming — one that required careful nesting and visual debugging.
Many early conference websites included embedded JavaScript for simple interactivity, such as image rollovers or form validation. This further blurred the line: a page that contained <script> tags was often described as “programmed in HTML,” even though the actual programming was in JavaScript. The conflation was harmless but persistent, and it appears in the informal notes and email archives of several PRASA working groups from that era.
From Static HTML to Dynamic Research Platforms
By the late 2000s, the limitations of static HTML became apparent. Conference proceedings with hundreds of papers required automated indexing, search functionality, and version control. Research groups began adopting server-side languages (PHP, Perl, Python) and databases to manage their online presence. Yet the transition was gradual. Many labs maintained a hybrid approach: a static HTML front page with links to dynamically generated paper listings.
For pattern recognition, the move toward dynamic platforms also meant that early HTML-based resources were often lost or archived only as PDF snapshots. The PRASA proceedings themselves shifted from HTML abstracts to full PDF collections, as seen in the digital archives from 2002 onward. A detailed look at the preservation of those early materials reveals how much context can be lost when the original HTML structure is abandoned.
- Common uses of HTML in mid-2000s pattern recognition labs:
- Conference proceedings pages with paper abstracts and links
- Lab member directories with research interests and photos
- Dataset documentation pages, often with download links
- Online demos using HTML forms and JavaScript
- Preprint repositories with manually updated index pages
A Concrete Example: The PRASA 2002 Proceedings
While the PRASA 2002 proceedings have been preserved as PDFs (as discussed in our earlier post Inside PRASA 2002: A PDF Time Capsule from Early 2000s Pattern Recognition), the original conference website was built entirely in HTML. The index page contained a table of contents with hyperlinks to individual paper abstracts, each formatted with <h3> and <p> tags. Researchers who attended the symposium recall that the HTML pages were hand-edited by a single volunteer, who had to ensure every link worked before the site went live. That volunteer, often a graduate student, effectively became the lab’s de facto webmaster — a role that required no programming skill, but plenty of patience with angle brackets.
This example illustrates how HTML, despite its simplicity, was a critical infrastructure for academic communication. Without it, the research presented at PRASA would have remained invisible to anyone outside the immediate conference attendees.
For those digging through the PRASA archives today, understanding the original HTML structure can help reconstruct the metadata that PDFs alone don’t capture — such as the ordering of presentations, the names of session chairs, and the exact wording of abstracts before they were reformatted for print. It is a small but valuable skill for any historian of pattern recognition research, and a reminder that even a markup language can leave a lasting footprint on the field.