Research institutions should treat maintenance of scientific software and code as an integral part of the research lifecycle rather than an optional afterthought. Sustained maintenance preserves reproducibility, protects institutional investment, and reduces wasted computational and human effort when code becomes unusable. Neil Chue Hong of the Software Sustainability Institute has long argued that software developed in research deserves the same stewardship as publications and data. Daniel S. Katz of the University of Illinois Urbana-Champaign emphasizes that treating software as a research output improves accountability and reuse.
Prioritize maintenance as core infrastructure
Funding mechanisms and promotion criteria should recognize maintenance work as valuable scholarly contribution. Grant proposals ought to include explicit maintenance plans, budgets for long-term support, and commitments to open licensing and archiving. The National Science Foundation and similar funders frame reproducibility and stewardship as expectations for publicly funded research; institutions can align local policy with those norms to create incentives for sustainable practice. Without these signals, maintenance tasks are often delegated to precarious positions or ignored entirely.
Invest in people, tools, and processes
Hiring dedicated research software engineers and creating stable positions for them turns maintenance into a staffed responsibility rather than a volunteer add-on. Complementary investments in continuous integration, automated testing, documentation standards, and community engagement reduce long-term upkeep costs. Training programs like Software Carpentry build researcher capacity to write maintainable code, which drives cultural change. Operationalizing maintenance requires predictable budgets and career pathways that reward stewardship activities.
Cultural and territorial considerations
Maintenance priorities must be sensitive to local contexts. Institutions in low-resource settings face different trade-offs than well-funded centers; capacity building, language-appropriate documentation, and regional collaboration networks can reduce inequality in software sustainability. Environmental consequences also matter: preserved, reusable code prevents redundant computation and lowers energy consumption across the research ecosystem. These nuances mean one-size-fits-all policies will underperform unless adapted to institutional mission and regional realities.
When institutions embed maintenance in governance, funding, and reward structures, they protect scientific value over time, increase public trust, and enable cumulative progress. Failure to act risks lost datasets, irreproducible results, and repeated reinvention that drains global research capacity.