2026 marks the 30th year since I got my first paying job in computing. I was still in college at the time, and so without much money, and decided to put some feelers out to a few department heads at my college to see if any professors wanted help with getting their course materials up onto something called the “World Wide Web.” Amazingly, there was one physics professor who reached out expressing interest and after meeting he agreed to pay me minimum wage1 to help him design and create a web-page for his courses.2 In a fit of nostalgia, I’ve decided to look back at the computing industry from that time period to see what it was like, and how it’s changed since. I may make this a regular feature, so I’ll likely touch on a lot of the nuance that I skip below. That said, I’ll just read and see what pops into my head.

Byte Magazine, January 1996

The January 1996 issue of Byte interestingly enough has a surprising amount of information on Linux, although I shouldn’t be too surprised since my school had a dedicated Linux lab that I used exclusively since it was usually empty. That said, most of the framing around Linux in this issue centers on its UNIX nature.3

One of the more interesting articles to look back on in this issue is titled “How the Web Will Change Computing” and is written by Steven Vaughan-Nicols and Rachel Schmutter. The article focuses a lot of attention on Netscape Navigator 2.0 and its latest feature to support the “dynamic interpreted” Java programming language. The thinking at the time was that the browser was the new operating system, and so Netscape was positioned as a platform to host an entire corporate software suite. Indeed, the professional version of the browser sold for $79 USD4 and provided WYSIWYG HTML authoring tools and a word-processor. As a budding software developer in 1996 I can recall that the “suite” model was ‘the future’ of software (Netscape, Lotus Notes, Delrina, etc.) and that model influenced my own thinking for a large chunk of my career. Interestingly, while Java is mentioned as a powerful extension to Netscape 2.0.5 The Netscape article was prescient on a number of points, and missed the mark on others.

The Datomic peer model is an interesting contravention to the ubiquity of thin clients. The January 1996 world viewed network bandwidth as the bottleneck and so it made sense (in hindsight) to predict that software should move to the server. However, Datomic is built with the assumption that network latency is the bottleneck instead and so (simplifying here) pulls a bunch of data to build local indexes, which allowing queries to run without the need for network hops. That said, even Datomic offers a client for applications that would benefit from it.

I want to talk a little bit more about the suites-view that infected the 1996 computing landscape and how it was congruous to a push for thin-clients. You see, at the time, pushing for software to adhere to a suites model led to a crisis of “bloatware” that brought even beefy desktop systems to a crawl.6 It was untenable to force customers to upgrade to top-end systems and companies were loath to cut features or offer lite versions of their software in fear of undercutting their own products. A promising solution seemed to be thin clients, since you could beef up the server systems to run bloatware, and the desktop only needed to draw pixels and chat over the network. This could work for some applications and in specialized networks, but thin clients were not a blanket solution because bandwidth was still a bottleneck in general in 1996, thin clients were also prone to crawling. So while my professor (and the industry) were right that think-clients were the future, they hit a wall in those early days because applications had to do a bunch of tricks to avoid chatty connections, which sacrificed interactivity. It was a mess. It wasn’t until the bottleneck switched from the size of the pipe to the speed of light (so to speak) that thin clients found widespread utility… it only too 20 years or so for infrastructure to get there.

This issue of Byte also had a cover story on “The World’s Fastest Computers” that had information about supercomputers, MFLOPS, parallel processing, and all of that garbage.7 These kinds of articles were crack to a younger me,8 but almost all of the actual gains in computing power since 1996 have been horizontal rather than vertical. Ironically, the article hypes up the future of super-powered general purpose computers, but doesn’t even realize that it’s looking at the tail-end of that paradigm.

So it goes.

F

At the time minimum wage was $4.25 an hour, which for someone taking in negative thousands of dallars a year seemed like a bonanza.↩︎ 1.

My development environment (my IDEEEEK so to speak) at the time was essentially a text editor and a prayer. There were no frameworks and no best practices and every bit of knowledge that I had was spelunked from “view source.”↩︎ 1.

At the time, the larger computing industry considered Linux (if they considered it at all) as a kind-of diet-UNIX, rather than as the monster that would eventually “diet” on the computing industry.↩︎ 1.

These days, the idea of paying $79 for a browser is inconceivable.↩︎ 1.

JavaScript is not mentioned at all! I’m certain that it would have been supported in 2.0, but mentioning it didn’t even rise to the level of afterthought at that time.↩︎ 1.

Every software box had a “System Requirements” box that, more often than not, were outright lies.↩︎ 1.

This wasn’t unique to Byte, so i don’t want to appear to pick on them. The whole job of reporting on computing, now and then, is to breathlessly report on what appears to be the pinnacle of computing’s trajectory, which is perpetually happening right now.↩︎ 1.

1996 me would have never imagined that I would own a supercomputer, and even carry it around in my pocket.↩︎