Penalty for not using Github PRs is massive

[kornelski]

The exponential penalty kills packages that have only 1 author. As I'm sole maintainer of some of my packages I never need to do PRs and I don't close issues, but I do still add new code, fix bugs and make new releases.

It's understandable that there should be some negative score for having bus factor == 1, but if someone doesn't embrace GitHub for issues and PRs their package will be ranked down into oblivion regardless of any other merits.

Suggested fixes:

• instead of months_since_last_pr_merged, use min(months_since_last_pr_merged, months_since_last_release) (or months_since_last_push), so that GitHub usage is irrelevant as long as the package is getting new releases.

• cap github inactivity penalty not to exceed sum of all positive scores from other github factors, e.g. score = max(0, subscribers + contributors_up_to_100/*etc*/ - months_since_last_pr_merged/*etc*/).

• the exponential could be smaller for packages with version >= 1.0, assuming the package may stable and doesn't need regular changes and fixes.

[MoSal]

but if someone doesn't embrace GitHub for issues and PRs their package will be ranked down into oblivion regardless of any other merits.

so that GitHub usage is irrelevant as long as the package is getting new releases.

Note that repo scores don't affect the order of results at all.

cap github inactivity penalty not to exceed sum of all positive scores from other github factors

That will make the factors useless as lengthy inactivity detectors.

the exponential could be smaller for packages with version >= 1.0, assuming the package may stable and doesn't need regular changes and fixes.

• Exponents and weights are definitely up for debate. And examples where the current ones don't work
  very well would be great.
• I'm against version values affecting the scores in any way.
• RE stable crates, I left a comment in Distinguishing between dead and stable packages #5.