[rb-general] Core Debian reproducibility: 57% and rising!

[beuc at beuc.net]

Hi,


On 29/10/2018 02:45, David A. Wheeler wrote:
> On Sun, 28 Oct 2018 17:42:58 -0700, Vagrant Cascadian <vagrant at debian.org> wrote:
>> And now I've found 88 binary packages!
> That's spectacular, and is a big tribute to the hard work everyone here is doing.
>
> I propose that the following go into the weekly announcements:
>
> * Vagrant Cascadian used a simple script and found that current Debian as delivered,
>    57% of the packages installed in a minimal system are verifiably reproducible
>    (88 reproducible and 66 unreproducible).  This is a major step forward,
>    as this is not a measure of special reproducible builds created "on the side", but
>    of the actual "real world" Debian packages including corroborating .buildinfo files.
>
> Does that capture things correctly?
>
> I'll try to add something like that in a little bit to the announcement file,
> but I fear I'm stating something completely wrong.

For the layman this 57% number sounds awfully low compared to other news
(e.g. tails says their ISO is 100% reproducible
https://tails.boum.org/contribute/reports/SponsorT/final/#index1h1 ;
FreeBSD announces 90%-100% from
https://reproducible-builds.org/events/berlin2016/2017lookahead/ )

I would skip the numbers or put them last in the news bit.
Or mention our non-real-world (higher) reproducibility percentage as
well, otherwise this is plain confusing, despite the TL;DR explaination
that follows.

Cheers!
Sylvain