Hacker News Reader

Comment by antihipocrat | Hacker News Reader

With so many developers using this package there must be a few here on HN. If anyone here has included this as a dependency I'd love to know the reasoning behind it.

View on HN

To prove a point that download counters are useless I wrote this: https://pypi.org/project/pypular/

Takes me about 7 minutes at home to trigger 1 million fake downloads.

I want to use http2 to send less headers and be able to do that in less time.

> download counters are useless

Unless they apply some kind of filtering to discount requests from the same source. If each source address were to only count as a single download once per period and a period was 20 minutes, your million fake downloads if from a single source would be forced to not take 38 years rather than 7 minutes.

They would not. In most cases downloads come from a badly setup CI that re-downloads the same thing tens of thousands of times per day. It's incredibly common for them to come from the same source.

It would make sense but ALL the project counters would go down by a lot.

You can fake counts up but not down. I think it can make a decent measure of popularity where there’s disincentive to be popular (like pypi security). You can set a point where you’ll catch almost all the truly popular packages, plus a small portion of the non-popular. (Although, the ratio of true positive to false positive may be way off, that’s the nature of things with low prevalence).

A large bulk of those downloads are coming from downstream packages (handlebars-helpers), I don’t think developers are pulling this in directly.

From the description of the repo on GitHub:

> I created this in 2014, the year I learned how to program. All of the downloads are from an old version of https://github.com/micromatch/micromatch.

Reply on news.ycombinator.com