Post

did you know that we correct known data anomalies in npmjs.com data? 👉 thanks to @jyc.dev 🩵

How is this done? Hopefully this is not reported data, otherwise can be used for abuse.

context

What prevents abuse? Maintainers of npmx are changing the data of their other projects, don't you see how this smells bad? Even if all is correct, relying on self-reporting means the incorrect data on other projects makes those look bad by comparison.

I’m fine with this as long as they publish who reported the anomaly and why. If it’s public it’s maybe more likey to be scrutinized. Also the ability to toggle between the sanitized and raw charts is reasonable. The question is, how maintainable is this?

Also clearly noted in the chart itself.

One concern I have is, will people start to use this as a means of “explaining” or rationalizing downloads of projects. Who ends up invigilating what an “anomaly” is considered and could that extend to classifying fair or unfair downloads? Does this leave room for litigious banter. That’ll be tricky

My solution is: be very, very clear about how what this process is for (anomalies) to mitigate litigiousness. Setup an approval process that is bulletproof.

This still relies on reporting so only a very small % of affected packages will have corrected data.

That’s the thing. If an anomaly is big enough to notice for the ecosystem (think the vite spike), it should be obvious enough to crowdsource.

There are over 3 million npm packages. Crowd sourcing skews towards the Very Online.

It's not reasonable for data to be skewed for some packages and not others. A user reading the pages isn't aware of the implementation details. But for this particular case, if there was a known issue with this date range, why not just exclude it from all packages?

Because it didn’t impact all packages. So you’d skew 99% of the data because of 1%. Doesn’t feel like that’s the solution

Sounds like a good reason to remove a flawed feature.