A version of this post originally appeared on Tedium, a twice-weekly newsletter that hunts for the end of the long tail.
A New Yorker and a tourist walk into the subway station at 42nd Street, also known as Times Square. One is delighted to be there; the other, extremely annoyed. One knows how to get out of there as quickly as possible. The other doesn’t. The New Yorker and the tourist are distinct but in this moment, they are the same.
Both are about to swipe their MetroCards, subject to the whims of the Metropolitan Transit Authority and the unheard-of reliability of a marginally successful operating system from the early 1990s.
And that operating system remains in use as the subway system feels the daily strain from heavy use.
According to one metric, 5.7 million people rode the New York City subway on an average weekday in 2016. This was the highest average total for the system since 1948.
If you ask an anecdotal average New Yorker, their response is usually, “That’s it?” Their disbelief is understandable since the city has some 8 million permanent residents and often swells to as much as 20 million during peak hours and holidays. (Guess a lot of people really enjoy flagging down cabs.)
But either way, 5.7 million is a lot, and one of the elements managing it is an operating system that fell into obscurity a quarter-century ago: IBM’s OS/2.
Betting on the future is hard, but the MTA kind of did it
In a Tedium article from this past March, Ernie Smith wrote about IBM’s big bet on a microkernel for operating systems—effectively a minimal layer of software through which an operating system’s basic tasks are handled—a bet that included a variant of its well-known-if-less-remembered OS/2. His article demonstrates how big the bet was. However, IBM’s confidence in its operating system prowess led others to take similar chances.
No bigger bet was made than the one by the MTA, the Metropolitan Transit Authority, who needed some process to eliminate tokens while moving into an era where everything was expected to be digitized. The result was the iconic MetroCard. A thin slice of yellow plastic with a prominent black strip, the MetroCard has been a staple of New Yorkers wallets since its introduction in 1993.
The story of how the current method for accessing New York’s subway is interesting for its insights into public infrastructure and the way it serves the public. Before we can touch that topic, it’s helpful to understand how the current system came to be. Because when you build something as important as the infrastructure to the NYC subway, it needs to work.
You pretty much get just one shot—and any mistakes will likely cost billions to repair and frustrate the lives of millions. Among many choices, one of the most robust turned out to be one of IBM’s most high-profile failures.
How IBM’s much-hyped but underwhelming OS found a home serving millions
Smith’s previous piece on the story behind IBM’s push into microkernels and its overall failure gives a whole host of backstory that I’m only going to briefly touch on here. For my purposes, the most relevant line from that article is nice and succinct: “OS/2, of course, did have its adherents.”
New York City's subway system was one of those adherents. The role of OS/2 in the NYC subway system is more of a conduit. It helps connect the various parts that people use with the parts they don’t.
“There are no user-facing applications for OS/2 anywhere in the system. OS/2 is mainly used as the interface between a sophisticated mainframe database and the simple computers used in subway and bus equipment for everyday use. As such, the OS/2 computers are just about everywhere in the system,” OS/2 and MTA consultant Neil Waldhauer said in an email interview. Waldhauer maintains an important OS/2-based tool that helps the MetroCard system function.
The reasons why the MTA ultimately decided to utilize OS/2 as it digitized some aspects of the subway system mirror the hype surrounding the operating system’s launch in the early 1990s. But a lot of that conversation and development started years before. Behind the scenes, Microsoft and IBM were working on the next generation of operating systems. While the general belief in popular culture dictates that Gates and Microsoft fleeced IBM over MS-DOS (the truth is more complicated), at the time IBM clearly felt it had a strong partner and joined forces with Microsoft again.
Rather than bemoan lost profits, IBM seemed to recognize a gap in its knowledge and it began the push to develop next-generation operating systems at a fundamental level, at first in partnership with Microsoft. This, almost predictably, worked out for IBM about as well as the MS-DOS deal did. However, in a very narrow window in the late 1980s, executives at the MTA were looking to remove tokens from the subway system and replace them with a prepaid card. The benefits were obvious, allowing for easier fare increases while offering tiered pricing. Riders would now have the option to choose between an individual or round trip option and an unlimited option that covers a set amount of time.
To implement this revolutionary upgrade, the MTA went with a known entity, IBM. At the time, it made a lot of sense.
“For a few years, you could bet your career on OS/2,” Waldhauer said.
To understand why, you need to look at the timing. “The design is from a time before either Linux or Windows was around. OS/2 would have seemed like a secure choice for the future,” Waldhauer said.
So for a lack of options, the MTA went with OS/2. And it’s worked out for decades, as one of the key software components of a complex system.
And it might even survive beyond that, according to Waldhauer: “I will go out on a limb and say that as long as MetroCard is accepted in the system, OS/2 will still be running.”
This is an exceptionally interesting point because the MTA is currently testing a system that can replace MetroCards with various forms of contactless payments. This transition should make things more efficient, while helping the MTA collect additional revenue.
It may sound nice, but the gaps are easy to see, especially if you look at a weird quirk in the current MetroCard system.
The mysterious magnetic strip and how it affects the lives of others
The move from tokens to MetroCards took years, and was by no description smooth. Tokens were officially phased out in 2003. By then, MetroCards were accepted at every station in the city—but no one was happy.
Access to the subway is usually easy but complaints about swiping your MetroCard are everywhere. And a lot of this seems to be a stupid disconnect between various parts of the system. While OS/2 is used to connect various parts of the subway system to a larger mainframe, the input components weren’t held to a higher standard.
The turnstiles in any given subway station in NYC are notoriously fickle—but they could work with IBM’s system.
Despite the failure of OS/2 in the consumer market, it was hilariously robust, leading to a long life in industrial and enterprise systems—with one other famous example being ATMs.,
“Thinking about all the operating systems in use [in the MTA], I’d have to say that OS/2 is probably the most robust part of the system, except for the mainframe,” Waldhauer said.
It’s still in use in the NYC subway system in 2019. IBM had long given up on it, even allowing another company, eComStation, to maintain the software starting in 2001 after it became clear that there was still a maintenance need. Eventually, another firm, Arca Noae ended up carrying the OS/2 torch: it sells an officially supported and regularly updated version of OS/2, ArcaOS, though most of its users are in similar situations to the MTA.
At this point, we’re talking about an OS designed in the late 80s, and released in the early 90s as part of a difficult relationship between two tech giants. The MTA had to ignore most of this because it had already made its decision, and changing course would cost a lot of money.
The integration between the backend and the things New Yorkers/tourists actually confront can be stupidly uncoordinated.
“I feel like the designers really considered MetroCard to be a mainframe database application with some random electronics to tie it together," Waldhauer said.
And now we get to talk about the magnetic strip. The black bar at the bottom of every MetroCard, regardless of branding, simply has to work. How it actually works is, for an obvious reason, a secret.
“People have hacked the MetroCard,” Waldhauer said. “If you have a way to see the magnetic encoding, the bits are so large you could see them under a magnifying glass. The encoding of the magnetic stripe is so secret that I have never seen it… It’s amazing the lengths people will go to for a free ride.”
Why does any of this matter? As a point, it really doesn’t. The MTA has made it clear it wants to move to contactless payments, just like the Oyster Card in London. But that process also has issues. It even hired Andy Byford, who spent years in leadership roles with the London and Toronto train systems, to help modernize the NYC Subway—with the end goal of ultimately eliminating MetroCards.
Access to the subway of the future will be contactless, using a patchwork series of devices and cards that will largely rely on digital payments. With device partners such as FitBit, catching a train in midtown will be more like queuing for a roller coaster at Disney World. If New Yorkers are lucky, the process will still easily accept cash but even that is no guarantee.
As I highlighted at the beginning of this, the MetroCard, for all its weaknesses and with dusty but functional technology at its core, was the great equalizer. But with a system that seemingly favors users with high-end phones, smart watches, and bank accounts, are we moving towards a world that feels a lot less equal? (And consider that we managed to ask these questions without even really diving into the Great Swipe Debate.)
It's an open question that we may only know the true answer to as the old tech gives way to new.