[Packaging] Distribute in Debian Repo

[asfimport]

Distribute Arrow in Debian repo: 

https://www.debian.org/doc/manuals/distribute-deb/distribute-deb.html#adding-packages-to-debian

 

Required to publish Arrow R package into CRAN.

Reporter: Javier Luraschi / @javierluraschi

_{Note: This issue was originally created as ARROW-4450. Please see the migration documentation for further details.}

[amoeba]

This is no longer a blocker for publishing to CRAN as the original issue says (arrow has been on CRAN for years) but including Arrow in the official Debian repositories would be beneficial to the R package, PyArrow, and other software that depends on Arrow C++ (notably GDAL, Ceph).

I'm not familiar with Debian packaging but there are some things that act in our favor as far as being able to get this done:

• Unoffical Debian packages are already published to JFrog and the work to do that is in this repo at https://github.com/apache/arrow/tree/main/dev/tasks/linux-packages/apache-arrow/debian. Users currently install the unofficial Debian packages this way.
• Members of the Debian Science Team are working on this already, with recent work in 2024:
  • Experimental repository for testing: https://salsa.debian.org/science-team/arrow
  • Discussion around this work: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=970021

I think a reasonable next step here is figuring out what issues the Debian Science team are running into and contributing to that effort.

[kou]

See also the past discussion on [email protected]: https://lists.apache.org/thread/4lfyt7nmfykg244fwhynpflv0m66hsvh

FYI: "official"/"unofficial" in this context mean that packages exist in deb.debian.org. The packages in https://apache.jfrog.io/ui/native/arrow/debian/ are the "official" packages in the ASF context but the "unofficial" packages in the Debian context.

[amoeba]

Thanks @kou!

One thing I pick up on from the discussion is that, in addition to the mechanics of creating and submitting a package to the official deb repos, there may be quite some work to adequately fill in the software licensing info. It all seems doable.