In the monthly meeting today, we discussed several larger efforts that are in the design stage. Currently these are:
cythonization of the scheduler - @jakirkham , @martindurant
refactoring the worker state machine - @fjetter, @crusaderky
configurable backends for dask collections (aka how to not import dask_cudf directly) - @rjzamora, @jacobtomlinson, @quasiben
Historically discussions about any size project have taken place in github issue, but there is a general sentiment that we want to try out a little bit of a heavier design process. To that end there is a new repo: https://github.com/dask/design-docs where people can create markdown files with design proposals. That way we can have conversations in PRs.
NOTE: This is definitely an experiment and we are still trying to avoid jumping to a full enhancement proposal process.
In the monthly meeting today, we discussed several larger efforts that are in the design stage. Currently these are:
dask_cudfdirectly) - @rjzamora, @jacobtomlinson, @quasibenHistorically discussions about any size project have taken place in github issue, but there is a general sentiment that we want to try out a little bit of a heavier design process. To that end there is a new repo: https://github.com/dask/design-docs where people can create markdown files with design proposals. That way we can have conversations in PRs.
NOTE: This is definitely an experiment and we are still trying to avoid jumping to a full enhancement proposal process.