remove looser relationships?

[NickCrews]

in #75 (comment), @PatMyron said:

For simpler regexes, I occasionally intentionally include corner cases like Billiam and Boberto? since I'm matching real datasets, so it's no issue that they aren't typical. Inversely, false positive matches are detrimental, so I exclude looser matches like:
bill,william,billy,robert,willie,fred
Curious if you'd consider excluding or separating looser matches!

Now I @NickCrews reply:

Yes, I would be willing to remove some looser matches. I would lean towards removing the "fred" from that list, but keeping the rest, as I know Bill/Robert is a moderately common solution.

Regardless, my goal is the for it to be trivial for users to customize this list to their needs, with both adding and removing name relationships. Then it isn't nearly as important for us all to agree on the level of strictness. Haave you found it easy to customize this list? Are you using the python API, or the raw csv, or what?

[PatMyron]

Just using the raw .csv so far! If the library distributes more conservative output, I (and others) can easily patch in custom data additions! Inversely, removing looser matches seemingly entails inspecting every new entry or ~pinning version(s) to ~fork from? (unless I'm missing better alternatives?) That asymmetry prompted wondering if the base package should err stricter

[NickCrews]

I think pinning versions is ALWAYS a good idea: look how in 9cc7712 I completely switched formats of the CSV, I'm sure that broke some people who weren't pinned.

Then, assuming you are working with a pinned version, removing rows should be just the tiniest bit harder than adding them.

[NickCrews]

Every time you bumped pinned versions, then yes you would need to manually look for changes that you didn't like that you might want to remove.

[PatMyron]

pinning versions is ALWAYS a good idea: look how in 9cc7712 I completely switched formats of the CSV, I'm sure that broke some people who weren't pinned.

Agree pinning helps with the caveat that producers and consumers can+should make ecosystems more robust. Even dependents following best practices appreciate stability + lower maintenance with how many dependencies most have nowadays

need to manually look for changes that you didn't like that you might want to remove.

yup, being stricter involving reviewing thousands of data lines is that asymmetry favoring stricter base packages