Bennett Cyphers

Custom methods are in, as of #60. Still need to do custom metrics.

Good points. There actually was architecture for this in earlier versions of ATM: if PCA was part of the pipeline, it would be computed first, then the intermediate data representation...

Actually, there should be two different functions: 1. one to delete a dataset or datarun and all associated data from the database, and 2. one to finish a datarun prematurely,...

@jtanman great suggestion. Yes, we can totally add that. More generally, there should be an API in `database.py` for amending/updating datarun info.

Yes! The plan is to eventually add support for other kinds of models, starting with regressions. However, this will come a little bit down the road -- if you'd like...

On second thought, maybe we should get CircleCI up and running first with the couple of tests we do have.

@jawz101 That's a cool suggestion, and I do think it would be neat to have something like that down the road. However, I don't think Privacy Badger is set up...

See also https://github.com/EFForg/privacybadger/issues/616#issuecomment-160812449

For reference, here's Privacy Possum's implementation: https://github.com/cowlicks/privacypossum/blob/4709d1da72bb142c2a08ec2ca1644e81bd0e808e/src/js/reasons/etag.js Their approach requires storing etag data after the first observation, so we'd need a pretty big new persistent data structure.

It looks like HotJar in particular uses third-party cookies, so it will be blocked if the user visits enough sites that use it. That doesn't mean we shouldn't try to...