There is no question that the world has a COBOL problem. Our banks, governments and enterprise systems are propped up by a 60-year-old programming language that fewer and fewer people understand.
Dragged under the spotlight by an incident in the US back in April, the COBOL problem is being dealt with as a simple skills shortage. As such, both IBM and Micro Focus (which play an outsized role in the upkeep of COBOL) have launched separate training initiatives, in a bid to incentivise developers to return to this no longer fashionable language.
However, it’s possible these firms have been wrong about the nature of the COBOL problem all along. According to technology start-up Phase Change Software, it’s easy enough for seasoned developers to learn COBOL, but it’s much harder to understand the language in the context of real-world applications.
“The market sees the COBOL problem as a skills shortage, but it’s actually a knowledge problem,” said Steve Brothers, COO at Phase Change.
“In the COBOL space, you have millions of lines of active code and, to perform necessary maintenance, you need developers that understand what that code does. But when you’re writing complex applications, code written in the morning becomes legacy by the afternoon.”
According to Brothers, as businesses suffer knowledge attrition (as developers retire or move on), the tendency is to solve problems by finding workarounds (i.e. building on top of and around existing code) instead of genuine solutions.
Eventually, this leads to bloated, unwieldy applications that even the most grizzled developers have no hope of understanding. This, he says, is the COBOL problem that merits the most urgent attention.
Addressing the COBOL problem
According to figures from Reuters, the majority of developers still familiar with COBOL are moving towards the twilight of their careers; only 11.5% are under the age of 35, while 18.8% are 55 or older. And the popularity of the language among budding devs has also been on a steep downward trajectory since circa 2002, falling from 5th to 33rd most popular programming language.
However, while addressing this imbalance is quite possibly the first step to rectifying the COBOL problem, new developers (no matter how talented) can only do so much to tweak legacy COBOL applications. They are hamstrung, somewhat, by a fear of breaking applications they do not (and cannot hope to) understand.
“The willingness among banks and insurers to change [COBOL] code is so low, because they see the risk is so high. They would rather pay fines for non-compliance than make changes to code that could break applications and crash their systems,” Brothers told us.
In a bid to address this sorry state of affairs, Phase Change will soon release an AI-based tool called COBOL Colleague, which the firm claims is able to direct developers towards the specific sections of code that require attention.
The tool is designed essentially to replace subject matter experts, of which there are increasingly few. In theory, new developers brought into an environment won’t need to rely on a steer from an experienced colleague (as is currently the case), but will be able to execute vital maintenance on the correct portion of the code base.
Modern developers spend up to 80% of their time identifying problematic code, says Brothers, gesturing to a piece of research conducted by Microsoft. A tool capable of slashing this time to discovery, then, could not only bypass the COBOL knowledge problem but yield dramatic productivity gains too.
A beta version of COBOL Colleague is currently in use at one bank in North America and talks are underway with another in Europe. The tool is scheduled for a full launch in Q2 of next year but, until then, we’ll all have to hope the glue and rubber bands holding together our banks and businesses holds fast.