README.md

Building Open Science Services on European E-Infrastructure (BOSSEE)

A proposal for INFRAEOSC-02-2019 started by participants in OpenDreamKit.

• brainstorm notes on hackmd
• Tentative aims

The draft is built automatically on CircleCI. To see a rendered draft, click on a successful build and then the 'Artifacts' tab, where there will be a link to the resulting PDF:

Or follow the URL on this page to view the latest build.

To compile document

install latex class (only required once)

git submodule init; git submodule update

to compile document

make draft

produces draft.pdf if all goes well.

writing

Most writing should occur by creating a task in tasks/ based on tasks/template.tex. Then the task should be added to a workpackage file in the task list.

outline

Outline of workpackages and tasks:

~11 FTEs

WP1 Administration

FTEs: 1

WP2 Structural improvements to Jupyter ("Enabling e-infrastructure")

FTEs: 2

1. Collaborating on notebooks
   1. Real-time collaboration development
   2. moving state of notebook into server,
   3. Exploring novel interaction environments (UPSud)
   4. (and more?)
2. Accessibility of Jupyter
   • Link to EINFRA-12-2017
3. Maintenance of core JupyterHub infrastructure
4. XEUS C++ kernels and infrastructure

WP3: Developing the Jupyter ecosystem

FTEs: 2

1. Extension of binder service:
   • entry portal where binder service can be selected
     • mybinder.org
     • EOSC hub
     • Amazon S3 (against payment)
     • local machine?
   • more improvements
     • infrastructure for archive of containers for reproducibility (EuXFEL use case)
2. Exploring sustainability of the service
3. For reproducibility
   • Repo2docker development (more data sources, supported platforms)
   • integration of nbval (or similar button) and nbdime for display into notebook (maybe a button “validate”)
   • display state of cells (up-to-date, out-of-date, maybe up-to-date)? // depends on server-side state
4. Widgets - important to be able to host GUI-like interfaces
   • In general, improving interactive experience
   • standard scientific visualisations in notebook (including vectors)
   • controllable through widgets
   • state saveable within the notebook
   • also in 3d (-> k3d?)
5. Jupyter based workflow for computational and data science
   • middle way between interactive and automatic task execution (nbparametrise)
   • use notebooks as scripts (papermill)
   • coordinate execution of notebooks
   • connect to running processes (possible when document state is in kernel?) to see life updates
   • computational environment (links to binderhub improvements) (repo2docker)
     • Like repo2docker, but not images (build conda environments for hpc, for instance)
     • given some specification, build me the right environment
     • preservation of this environment
       • will it still work in 10 years time
       • [Following notes can be re-used in proposal to explain motivation]
       • Min: longevity of conda environments is short
         • most python packages are packaged per python version
           • pip does source install fallback
           • conda(-forge) relies on the the right python version
             • not building for Python 3.5 anymore.

WP4 Applications (Use cases? Pilot services?)

FTEs: 3

1. XFEL Open Science (LIGO OSC for XFEL?)
2. Personal data (Paul)
3. Maths demonstrator

at least one or two more

WP5 Services / EOSC integration

FTEs: 2

1. Prototype/operate European Binder
2. Migrate service into EOSC
3. Cooperation with EOSC

WP6 Education and dissemination

FTEs: 2

1. Training workshops
   1. N / year
   2. Some dedicated to diversity (WISE, etc.)
   3. Workshop surveys
2. Online training resources (competition issue?)
   1. Be more specific to our tooling (conda/python/jupyter/etc.)
   2. Generic resources like OpenAIRE, carpentry exist already
   3. Focus on gaps filled by our expertise; complementary, not competitive
3. Presentations at conferences to meet researchers where they are (maybe not task in its own right)
4. help desk (Nicolas)