Dave Hirschfeld

I suppose it's theoretically possible to rewrite history to add a `Signed-off-by` to a merge commit, but I don't think that really makes much sense and it would be better...

The DCO check does appear to be broken though: - https://github.com/databricks/databricks-sql-python/issues/520

I appear to also be running into this issue.

As someone responsible for specc'ing up and deploying an analytics environment I think this is a pretty important issue. If `cytoolz` isn't in your environment your code will just run...

...or just add a `'perf'` section to [`extra_requires`](https://github.com/dask/dask/blob/master/setup.py#L10-L17) - e.g. ```python extras_require = { 'array': ['numpy >= 1.11.0', 'toolz >= 0.7.3'], 'bag': ['cloudpickle >= 0.2.1', 'toolz >= 0.7.3', 'partd >=...

Note: this also applies to city/murmur hash... which aren't installed by default with either conda or pip. https://github.com/dask/dask/blob/334bc7fbdf20f6656ab4f6ef292ee926766db047/dask/hashing.py#L12-L16 https://github.com/dask/dask/blob/df7db48145b5a61e2f5fab7aa89da3256c8877e3/dask/array/core.py#L2160-L2165

Also, `scipy` - #3309 From my point of view - I would really like these dependencies listed somewhere, *especially* where they impact performance. There seems to be some pushback to...

I would use comments to split the `optional-requirements.txt` file into sections - those deps which will improve performance and those which are only required to interact with 3rd party libraries.

This appears to be the approach taken by `pandas`: https://github.com/pandas-dev/pandas/blob/master/ci/requirements-optional-conda.txt ...by default install just the minimal deps but list (somewhere) all optional deps, including any version requirements.

Unless there's a smarter way to handle this, perhaps type validation needs to be switched off for interpolated values?