Fellow hacker Simon Tournier did an experiment: reproducing a pure #Guix 2020 paper in the worst-case scenario—shutting down network access except for #SoftwareHeritage and #Disarchive, and from there rebuilding everything from source.
https://simon.tournier.info/posts/2023-06-23-hackathon-repro.html

Spoiler alert: there are lots of bumps on the road. But it works in the end and there are many lessons we can learn to improve on it.

#ReproducibleResearch #OpenScience

@wilfredh In #Guix, we went to great lengths to tackle disappearing source files with #SoftwareHeritage and #Disarchive:

https://guix.gnu.org/en/blog/2019/connecting-reproducible-deployment-to-a-long-term-source-code-archive/
https://hpc.guix.info/blog/2022/02/guix-hpc-activity-report-2021/

Timothy Sample just published an updated “Preservation of Guix” report:
https://ngyro.com/pog-reports/latest/

Key takeaway is that 87% of the source code of packages present in #Guix today is archived via #SoftwareHeritage + #Disarchive, 77% if we look at all the packages that appeared in Guix between May 2019 and today.

The database at https://disarchive.guix.gnu.org now includes #Disarchive metadata for tar.xz archives!

👉 https://lists.gnu.org/archive/html/guix-devel/2022-08/msg00061.html

Concretely, it means that .xz tarballs #Guix refers to can be recovered using that database + #SoftwareHeritage.

Timothy Sample and the SWH team discuss ways to integrate the #Disarchive database in #SoftwareHeritage:
https://sympa.inria.fr/sympa/arc/swh-devel/2022-02/msg00007.html

Good news for long-term source code preservation and support in #Guix!
👉 https://www.softwareheritage.org/2022/01/13/preserving-source-code-archive-files/

This grant will allow Timothy to further work on #Disarchive, the missing link between #SoftwareHeritage and the deployment tools such as #Guix. 👍

Timothy Sample published a new Preservation of #Guix Report, showing more details as to what’s archived and what’s missing from #SoftwareHeritage:
https://ngyro.com/pog-reports/2021-12-06/

The breakdown now shows tar.xz, tar.bz2, git, hg, svn, etc., which helps understand what’s not being archived or what’s not correctly handled by #Disarchive.

@kirschwipfel In a nutshell, SWH does not archive tarballs—it archives the *contents* of tarballs. #Disarchive is a tool to “disassemble” and “reassemble” tarballs so we can reconstruct them:
https://ngyro.com/software/disarchive.html

Slides and maybe video should be available soon…

Just finished presenting how #SoftwareHeritage helps #Guix achieve reproducible deployment, including the latest advances around #Disarchive.

Great to have the opportunity to share with this fine audience!

And of course, 👍 to Timothy, Simon, and everyone who made it happen in the first place. 🙂

Timothy Sample on tradeoffs in #Disarchive for lossless #tar header compression:
https://sympa.inria.fr/sympa/arc/swh-devel/2021-10/msg00044.html

#Guix now has its own infrastructure to build a #Disarchive database, which, coupled with #SoftwareHeritage, is our guarantee for long-term source code availability:
https://lists.gnu.org/archive/html/guix-devel/2021-10/msg00080.html

Good news! New #Disarchive release makes it into #GNU #Guix!
https://git.ngyro.com/disarchive 🠴

The plan is to use Disarchive + #SoftwareHeritage as a fallback when downloading tarballs:
https://issues.guix.gnu.org/47336

👍 Timothy!

Recovering actual tarballs from content at #SWHeritage with #Disarchive:
https://forge.softwareheritage.org/T2430#47485

👍 to Timothy Sample who tackled this while some of us were sunbathing:
https://issues.guix.gnu.org/42162#15

Another example of the commitment of the #Guix folks!