Round Table Introduction: From an institution-developed package to a sustainable project
1. Round table:
From an institution-developed package
to a sustainable project
Turning institution-developed codes into
sustainable community-driven
projects
Mario Juric <mjuric@astro.washington.edu>
UW Astronomy | DIRAC | eScience
@mjuric
2. @mjuric
The Problem
> To solve their data processing problems, big science projects are
making significant $ investments and advances in analysis
software. Software is becoming as expensive as physical
instruments.
> These projects don’t have the resources or mandate to fully
publicize, generalize, or make their codes sufficiently open and
friendly for others to use or build on. The code effectively becomes
project-specific instrumentation.
> Issue: The the community doesn’t benefit as much as it could.
Others have to reinvent the wheel; the use of inefficient or
suboptimal methods continues.
3. @mjuric
Example: Large Synoptic Survey
Telescope
> LSST will process >100x
data than any optical
astronomy project before it.
> Software funded at ~M$50 level.
> Processes images to identify objects and measure their properties.
Likely to have the best (most accurate, most optimized) data
processing algorithms available. Already 95% general-purpose
(can run on different instruments)!
> Issues: Complex install (bespoke install system). Difficult to use.
Not sufficiently documented (a.t.m.). Few resources (people time!)
within the projects to improve those aspects.
4. @mjuric
Topics for Discussion
> We’d like to find a way to “liberate” the value that’s in this
codebase:
– Make it usable enough for other telescopes to try (and adopt)
– Make it usable so individual astronomers can benefit from improved algorithms
– Build a self-sustaining community of users and contributors, beyond LSST
> How would one take existing institution- or grant-developed
codebase and turn it into a sustainable community project?
> Can one even build a developer community around a specialist,
“company”, project? What are the experiences of projects that
have succeeded at building developer communities?
– Experiences that worry us: Netscape, OpenSolaris, AOSP, …
5. @mjuric
A Concrete Hack:
Use M$0.1 to leverage M$50+
> Assuming we have k$100-200 available, how should
we invest those $$ to have the largest impact wrt. the
goals above.
– What do we do?
> Documentation, packaging, outreach, support…?
– How do we do it?
> Give in to the “Hire a postdoc/graduate student” instinct?
> Or are there better ways?
> “It’s unlikely to work, don’t even try” is a valid answer!