Intentional OSS in science:
because OSS ≠ “one size”

Sidney Bell - CZI Foundation, Science Initiative

@sidneymbell (everywhere)

October 2023

Agenda

• Motivation & goals

• Overview of framework components

• Overview of draft archetypes

• Next steps & discussion

Motivations

“There isn't a single model there. You have to sort of say, ‘What's the right interaction mode with the community you're dealing with and what supports your objectives?”

"I don't think [people] have a very good perspective on what it even means... Just because you go build in a public GitHub repo and slap a MIT or Apache license on it..it's not the same thing as when people talk about open source being a force in the world."

Quirks of developing scientific OSS

"So, they [Python, Linux] somehow built a movement appropriate for the size of their problem. And so to me, I don't look at a lot of the stuff in science, which is always going to be by its very nature, kind of very small scale and nichey and never get a lot of momentum, both because of the size of the ecosystem and the use cases, but also just because it obsoletes itself very quickly. I see that as very distinct from the big platform building open source world. It's just a completely different thing."

Framework target outcomes

• Teams can articulate different models of OSS development in science and which one makes sense for their strategic goals
  
  

• Teams have the strategic guidance they need to incorporate their OSS model(s) into their roadmaps & long-term sustainability plans
  
  

• Teams start to integrate the most relevant OSS
  best practices for them into their workflows

How to use this framework

1. OSS Archetypes

• Based on the Mozilla archetypes, adapted for scientific software

• See the full matrix here

2. Self Assessment

• We use the OSS Needs Assessment because it is clear, actionable and pragmatic.
   

• There are no wrong answers. A higher score is not always “better.”
   

• Holistic evaluation of the alignment between your goals & your practices.
   

2. Self Assessment

3. Growth Plan

SWOT analysis workshop
(strengths, weaknesses, opportunities, threats)

Deep dive:

Scientific OSS archetypes

Feedback requested! Especially:

- Clarify and differentiation

- Practices

- Examples

- Metrics

https://tinyurl.com/oss-archetypes

Archetypes overview: institutionally supported projects

Archetypes overview:
more flexible project types

Scientific Platforms

Hosted applications with a single canonical instance

• Strategic goals:

  Democratizing access to analysis tools / data (often via a centralized GUI), which may have a standard setting effect for a field.

• Key practices:
  • Code accessibility (trust)
  • Licenses & copyright
  • Quality
  • User friendliness
• Risks: Sustainability, alignment with community
• Decision making: Host institution
• Execution: Host institution
• Markers of success: Adoption, quality

Nextstrain (web)

UCSC Genome Browser
Genbank

Rocket ship to the moon

Prototypes with rapid iteration by the founder

• Strategic goals:

  Fast and focused development to test a scientific or technological hypothesis.

• Key practices:
  • Code accessibility
  • Licenses & copyright
  •  Transparency & expectation management
• Risks:
  • Everything depends on success of original vision
  • Missing out on partnerships
• Decision making: Founder(s)
• Execution: Founder(s)
• Markers of success:
  • Speed of development
  • Achieving original technical goals and testing hypothesis

Seurat (early days)

Rocket ship to Alpha Centauri

More mature projects still owned by the founder

• Strategic goals:

  Fast and focused development of a high-confidence solution to a high-priority problem. Founders retain strong influence.

• Key practices:
  • + User friendliness
  • + Quality
  • (and all from Rocket Ship to the Moon)
• Risks:
  • + Low “bus factor” is a concern for long-term sustainability
  • (and all from Rocket Ship to the Moon)
• Markers of success:
  • + Adoption
  • + "Standard setting"

Seurat (now)

Scientific library / pipeline

More mature specialized project where any expert may contribute

• Strategic goals:

  Pool development efforts to solve shared core problems & standardize analysis.

• Key practices:
  • Transparency & Consensus building

  • Governance

  • Contributor community management
  • Quality & Releases
• Risks:
  • Small developer pool & high maintenance costs in rapidly evolving scientific fields
• Decision making: Core developer-driven
• Execution:  Open to anyone with domain expertise
• Markers of success:
  • User adoption, contribution quality
  • Code quality

Vispy

Wide Open

Democratic projects open to anyone

• Strategic goals:

  Large-scale collaboration; community can become self-sustaining

• Key practices:
  • Transparency & Consensus building
  • Governance
  • DEI & Contributor community management
  • Quality & Releases
• Risks: founders' willingness to share decision making; high maintenance of onboarding pathways
• Decision making: Group-based, consensus / democratic
• Execution:  Open to anyone
• Markers of success:
  • Contributor onboarding efficiency &
    community growth
  • Diversity of voices in decision making

Pandas, numpy

Acknowledgements

Questions?

@sidneymbell

https://sidneymbell.science