The blog is missing the link of the page. The awesome folks over at debian, mainly Holger Levsen, is helping Arch Linux achieve this with their infrastructure. The progress and status can be seen here: https://tests.reproducible-builds.org/archlinux/archlinux.ht...

Currently at 76% of 70% built packages.

for comparison, debian stretch on amd64 is at 94.1% of 24821 packages. glad to see the knowledge is being shared.

stats: https://tests.reproducible-builds.org/debian/reproducible.ht...

We work together with a few Debian developers involved with reproducible builds in the reproducible-builds project, so yes a lot of work is shared which is awesome :)

If it's a random sample, that seems to be a reasonable extrapolation (of course standard caveats apply).

http://a.co/a8yEzzt A Manga Guide to Statistics. No I'm not being snarky, it's a great read and I highly recommend it. Sampling, why it works and the philosophy behind statistics are all a bit unintuitive and so a light read is perfect for helping you develop that intuition.

I have a degree in mathematics, I'm fine on the statistics front. It's very likely to NOT be a random sample - you would only be testing the packages which have had effort put in to make them reproducible, for reproducibility. I would expect the non-sampled packages to be 0% reproducible, since the reason they were not tested is that nobody is trying to make them reproducible (yet).

>A Manga Guide

No thanks.

Well I didn't mean to claim it's 100% reproducible in the title. But we finally started working towards reproducible builds, we still need a way for users to easily reproduce builds, publish the results somewhere. And test with more differences, luckily Debian has done a lot of effort in this regard!