[Paper Review] A sustainable infrastructure concept for improved accessibility, reusability, and archival of research software

Timo Koch, Dennis Gläser+7 more|arXiv (Cornell University)|Jan 27, 2023

Scientific Computing and Data ManagementDecision Sciences3 citations

TL;DR

This paper proposes a sustainable infrastructure for research software that enhances accessibility, reusability, and archival through a multi-modal representation of software artifacts—such as containerized environments, browser-based apps, and source code—supported by FAIR-compliant metadata and open-access repositories. The approach enables practical reproducibility and long-term reuse by aligning with software engineering best practices and integrating with existing research workflows.

ABSTRACT

Research software is an integral part of most research today and it is widely accepted that research software artifacts should be accessible and reproducible. However, the sustainable archival of research software artifacts is an ongoing effort. We identify research software artifacts as snapshots of the current state of research and an integral part of a sustainable cycle of software development, research, and publication. We develop requirements and recommendations to improve the archival, access, and reuse of research software artifacts based on installable, configurable, extensible research software, and sustainable public open-access infrastructure. The described goal is to enable the reuse and exploration of research software beyond published research results, in parallel with reproducibility efforts, and in line with the FAIR principles for data and software. Research software artifacts can be reused in varying scenarios. To this end, we design a multi-modal representation concept supporting multiple reuse scenarios. We identify types of research software artifacts that can be viewed as different modes of the same software-based research result, for example, installation-free configurable browser-based apps to containerized environments, descriptions in journal publications and software documentation, or source code with installation instructions. We discuss how the sustainability and reuse of research software are enhanced or enabled by a suitable archive infrastructure. Finally, at the example of a pilot project at the University of Stuttgart, Germany -- a collaborative effort between research software developers and infrastructure providers -- we outline practical challenges and experiences

Motivation & Objective

• Address the challenge of sustainable archival and reuse of research software artifacts, which are often inaccessible or non-reproducible over time.
• Improve practical reproducibility by enabling researchers to explore and reuse software beyond published results, even when original results cannot be reproduced.
• Develop a framework that supports multiple reuse scenarios through diverse representations of the same software artifact, such as source code, documentation, and interactive environments.
• Align software archival with FAIR principles (findable, accessible, interoperable, reusable) to ensure long-term sustainability and utility.
• Integrate infrastructure solutions with existing research software development and publication workflows to minimize disruption and maximize adoption.

Proposed method

• Design a multi-modal representation concept that treats research software artifacts as snapshots of research states, supporting various reuse modes: installable code, containerized environments, browser-based apps, and documentation.
• Integrate FAIR principles into metadata and software descriptions to ensure findability, accessibility, and interoperability across repositories and platforms.
• Leverage public, open-access infrastructure such as Software Heritage and Zenodo for long-term archival and persistent identifiers (DOIs, URLs) to ensure stability.
• Adapt existing software installation workflows (e.g., git clone, pip install) to be compatible with long-term data archive access patterns, especially for high-performance computing and complex dependencies.
• Implement a decentralized, scalable model for interactive reuse, where users can contribute computing resources, reducing reliance on centralized, costly cloud infrastructure.
• Collaborate with research software engineers and infrastructure providers to prototype and test the system in real-world settings, such as the DuMu x and ViPLab projects at the University of Stuttgart.

Experimental results

Research questions

• RQ1 How can research software artifacts be archived in a way that ensures long-term accessibility and reusability beyond the scope of published results?
• RQ2 What multi-modal representations of software artifacts best support diverse reuse scenarios, including exploration, modification, and deployment?
• RQ3 How can FAIR principles be practically applied to software metadata and documentation to enhance findability and interoperability?
• RQ4 What infrastructure models enable sustainable, scalable, and decentralized reuse of research software while minimizing dependency on proprietary or costly cloud services?
• RQ5 What practical challenges arise when integrating archival and reuse infrastructure into existing research software development and publication workflows?

Key findings

• Approximately 25% of URLs cited in research papers between 2005 and 2017 were no longer accessible, highlighting the urgency of sustainable software archival.
• Only half of the 98 randomly selected bioinformatics software packages in cited papers were labeled as 'easy install,' indicating a strong correlation between installability and citation impact.
• The multi-modal representation concept successfully supports diverse reuse scenarios, including interactive exploration via browser-based apps and reproducible execution via containerized environments.
• Integration of software artifacts with public, open-access archives like Software Heritage and Zenodo enables persistent, FAIR-compliant access and versioning.
• Practical challenges in workflow adaptation—such as aligning standard software installation (e.g., pip install) with archival API access—require careful interface design and tooling support.
• A decentralized, community-driven model for interactive reuse is feasible and scalable, reducing dependency on centralized, costly infrastructure while maintaining compliance with FAIR principles.