Legacy Porting

Legacy Porting

Legacy porting refers to transferring old, often decades-old computer programs onto modern hardware or into a newer programming language. Because such programs frequently handle central tasks in banks, government agencies, and insurance companies, this overhaul is considered risky, expensive, and lengthy.

Many banks, insurance companies, and government agencies work with programs that were written in the 1970s or 1980s. These old programs often still run today because they work reliably and no one dares to touch them. They are called legacy systems. Legacy porting means: an old program like this is transferred onto new technology, without anything supposed to change for the users. This can mean that the program’s source code is translated into a more modern programming language. It can also mean that the program moves from an old mainframe to standard servers or into a data center on the internet.

Why nobody simply throws away the old code

Old software is not a problem because it is bad. It is a problem because fewer and fewer people understand it. The programming language COBOL, for example, still controls a large part of global payment transactions today. But the developers who master it are gradually retiring. Anyone who can no longer find someone able to maintain a system has a risk on their hands that grows every year.

Then there are the costs. Old mainframes require expensive maintenance contracts and special replacement parts. New features can only be added with great effort. A bank that wants to offer an app with instant transfers quickly runs up against the limits of a system from before the internet era. Experts then speak of technical debt: modernization has been postponed for years, and the bill comes due with interest.

On the other hand, such projects strikingly often fail. There are well-known cases in which government agencies and banks have spent hundreds of millions of euros and ultimately ended up back with the old system. It is precisely this mix of high pressure and high risk that makes legacy porting a recurring topic in the business press.

From the mainframe to the present: the usual paths

The simplest path is called rehosting. Here, the program remains unchanged in content; it merely moves to new hardware. This is comparatively fast, but it does not solve the fundamental problem: the hard-to-understand source code continues to exist. The most elaborate path is a complete rewrite. The software is written entirely from scratch, using the old system only as a reference.

In between lies actual porting: the existing source code is translated line by line into a modern language such as Java. Tools handle the mechanical part, while humans review the result. The catch lies in the details. Old programs often contain rules that are documented nowhere, because their authors considered them self-evident. Such hidden rules only come to light when, after the switch, wrong amounts are suddenly calculated.

For some years now, language models have been used for this purpose, meaning AI systems that can process text and program code. They suggest translations and explain what an incomprehensible section is presumably doing. However, they cannot be relied upon, since such models occasionally invent plausible-sounding nonsense. That is why, in the end, there is always a comparison test: the old and new systems calculate the same cases, and the results must match exactly.

Where the topic shows up in the news

Legacy porting usually becomes visible when it goes wrong. When a bank switches over its account system on the weekend and no transfer works on Monday, a migration is almost always behind it. Reports about delayed tax software or failed railway ticketing systems also fall into this category.

For investors, the topic is interesting because large IT service providers generate significant revenue from it. Companies like IBM, Accenture, or Capgemini have been earning money from modernization projects for years. At the same time, AI providers are promoting the idea that they can make such projects considerably cheaper in the future. Whether that is true remains an open question: translating the source code is only part of the work, testing and safeguarding usually costs more.

A related term that often comes up in the same context is cloud migration. This refers to moving to third-party data centers that are used over the internet. Legacy porting is the broader term and includes the language switch as well. Anyone who confuses the two quickly underestimates how much work actually lies behind a modernization.

Subscribe free. Unsubscribe the second it sucks.

High-signal news across AI, business, UX, and tech. Every morning.