CSE legacy code can involve several years or even decades of development and resources invested in it. Maintaining such legacy code comes with several engineering challenges and its imperative to know how to balance the critical goals associated with the code.
Successful computational engineering tools involve many person years of development. Nonetheless, software development costs alone represent only a portion of stakeholders' (developers, users, consumers and sponsors) total investment in such tools. There are many ways in which stakeholders become invested in a computational engineering tool. These include
- Software development costs
- User training and development costs
- Input database development costs
- Community adoption and development costs
- Perceived or measured commitment to validity of results
Eventually, investments in a computational engineering tool, both tangible and intangible, become so great that wholesale re-write (that is throwing existing code out and re-writing from scratch), is not seen as an acceptable strategy to move the code forward with new generations of computing technology.
The key software engineering challenge of legacy code involves balancing several goals
- Maintaining the original source code largely intact
- Maintaining acceptable performance with newer computing technologies
- Maintaining acceptable consistency in numerical results
- Maintaining external interfaces (both data and user)
To achieve these goals, changes to legacy code are typically evolutionary rather than revolutionary in nature. Eventually, however, even evolutionary code changes can become cost-prohibitive. Typically, by the time legacy code faces this demise, which could easily involve more than 30 years of development, the evolutionary approach has provided ample time for suitable alternatives to develop and become adopted.
