Balisage Paper: XML as a Technological Pinnacle

Introduction

As XML approaches 30 [XML 1996], and its parent technology approaches 60 [Goldfarb 1996], the peaks and valleys of its existence keep getting clearer. XML soared to great heights in its first few years, dipped during the dot-com bust, climbed again for a burst of AJAX and REST API use, and then faded again after JSON took much of its audience. Like many mountains, though, technological pinnacles leave behind powerful remains that keep marking the landscape long after they stopped being the biggest thing around.

What is a pinnacle?

A pinnacle is generally a more exciting word for a peak, often reflecting slightly more extreme circumstances. In architecture, for example, roofs can have peaks, but pinnacles are usually towers extending above the roof. Pinnacles are often, though not always, seen as less secure places to be than peaks. Pinnacles are often admired, seen as the culmination or greatest expression of something [Pinnacle Definition]. Annual Pinnacle Awards, for example, are meant to honor companies, products, and individuals that have demonstrated exceptional innovation, creativity, and impact in their fields [Pinnacle Awards]. As the annual nature of that award suggests, a technology pinnacle is often a temporary state.

On the peaks

Unlike many technologies, XML had two huge moments. The first peak was early, when XML was going to transform the Web, ecommerce, and everything [Bosak and Bray 1999]. The second peak arrived when XML thrived on REST APIs and its association with Ajax [Garrett 2005], before JSON took its data serialization crown away. Google Books' Ngram Viewer [Ngram-XML] provides a rough presentation of this pattern, shown in Figure 1, though two caveats are important. First, book publication often adds a year or two year delay. Second, the Ngram viewer is reporting the frequency of occurence of XML in books. If a book simply mentions Asynchronous JavaScript and XML to expand Ajax, or includes code using XMLHTTPRequest with any kind of payload, that will set off the counter. The hits in the first peak are likely more solid.

Figure 1: Ngram for XML, 1995-2022

Google Books shows two peaks for the use of XML, in 2002 and 2008, followed by a long slow fade.

While Ngrams are fairly blunt instruments, two additional searches can clarify what was happening in each peak. Both have many fewer appearances than XML, but they do show intersecting trends. XSLT, shown in Figure 2, was strongly associated with the first story, and Ajax, in Figure 3, with the second.

Figure 2: Ngram for XSLT, 1995-2022

Google Books shows one peak for the use of XSLT, in 2002, followed by a bumpy fade.

Figure 3: Ngram for Ajax, 1995-2022

Google Books shows a double peak for the use of Ajax, across 2007-11, really only dropping in 2016.

All of these technologies are still in use, of course, but all of them have descended from their peaks.

Pinnacles after the peak

Pinnacle technologies have a habit of lasting much longer than anyone expects, often even much longer than anyone especially wants. (COBOL is a legendary case, while Java is a more ordinary complaint.) Even in software, where the costs of change seem like they should be much smaller, technologies that had a good run can linger apparently forever. Data has even lower transition costs because of ubiquitous transformation tools, but people and companies hold on to their technologies as long as there remains enough community to support them.

One classic way that old pinnacles stay relevant is by staying close to related technologies. The Java Virtual Machine (JVM), for example, is used for languages far beyond the original language. Kotlin, Scala, Groovy, and Clojure are all built around the JVM, and dozens of other languages have implementations that use it [Urma 2014].

XML does not have this kind of a posse. In fact, the early design decision that XML documents must be well-formed makes it difficult to reuse XML tools on SGML, HTML, or a variety of other formats. However, a different approach, through techniques like Invisible XML [Invisible XML], makes it possible to bring XML tooling and transformations to other formats. XML itself remains largely unbeaten in applications that need the flexibility it provides, especially for custom structures and mixed content.

Just as regular expressions have remained themselves through decades of use in multiple environments, so XML will likely hold on. Regular expression tooling has drilled deeper into programming languages than XML parsers and other tools, so hasn't had to deal with as much turmoil [Posch 2025]. XML's early efforts to join the world of the Web browser got XSLT 1.0 into the browser, but even that is fading now [Freed 2025]. XML 1.0 itself seems safe there for now, too tightly bound into APIs and other processing, its notable naming of XMLHTTPRequest [XMLHttpRequest] still respected as legacy.

So which will it be? Will XML end up an isolated monolith like COBOL, requiring dedicated support while managers plot its long-term removal, or will it remain so integral to so many technologies that it lasts like the JVM?