[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[task #16228] Guix-Maneage synergy
From: |
Boud Roukema |
Subject: |
[task #16228] Guix-Maneage synergy |
Date: |
Tue, 12 Jul 2022 19:56:25 -0400 (EDT) |
URL:
<https://savannah.nongnu.org/task/?16228>
Summary: Guix-Maneage synergy
Project: Maneage
Submitter: boud
Submitted: Tue 12 Jul 2022 11:56:23 PM UTC
Should Start On: Tue 12 Jul 2022 12:00:00 AM UTC
Should be Finished on: Tue 12 Jul 2022 12:00:00 AM UTC
Category: None
Priority: 5 - Normal
Status: None
Privacy: Public
Percent Complete: 0%
Assigned to: None
Open/Closed: Open
Discussion Lock: Any
Effort: 0.00
_______________________________________________________
Follow-up Comments:
-------------------------------------------------------
Date: Tue 12 Jul 2022 11:56:23 PM UTC By: Boud Roukema <boud>
How can Guix structure or tools be useful to Maneage?
Comments about GNU Guix (and Nix) are currently in Appendix A.B.3 'Nix or GNU
Guix' (p3, [1]) of Akhlaghi+2021. The two main concerns are:
(1) Guix needs root privileges (violating our 'completeness' criterion). This
is related to the differing aims of the reproducibility of software binaries
versus the reproducibility of scientific results.
(2) With Guix, the user has to remember to store the Guix commit ID (and names
and versions and commit IDs of the various scientific software packages?), as
per Konrad Hinsen's 2020-01-14 blog [2] that we put as our footnote 11, while
in Maneage, this is built-in to the whole structure, and the user would have
to opt out from version IDs and sha512sums rather than opt in.
Initial discussions among speakers at the upcoming Sep 2022 Guix meeting [3]
suggest that:
(1)
(1.a) This is a problem of the host GNU/Linux OS, not of Guix itself, which in
principle could become user-level only, especially on GNU/Hurd.
(1.b) Guix could be used to produce a container or pack that satisfies our
eight criteria [4].
(2) Konrad will present more details in September.
My responses:
(1.a) This appears to be a long-term possibility, not practical currently on
many real OSes.
(1.b) If the container or pack does not contain the Guix system itself, then
it would seem to violate criterion 1 of being self-contained; a user of the
container or pack would not have the full knowledge of how the original
authors chose/patched/developed the software and
designed/implemented/debugged/revised the scientific analysis of the research
project.
(2) It would be best to see some working examples (preferably peer-reviewed)
and then see if they satisfy the eight criteria (apart from (1) needing root
access). These could be compared with recent Maneage papers listed under
'Published works using Maneage' in the README-hacking.md [5].
[1] https://ieeexplore.ieee.org/ielx7/5992/9459454/9403875/supp1-3072860.pdf
[2] https://guix.gnu.org/en/blog/2020/reproducible-computations-with-guix
[3] https://10years.guix.gnu.org/program
[4] https://doi.org/10.1109/MCSE.2021.3072860
[5] https://git.maneage.org/project.git/tree/README-hacking.md%23n172
or clickable:
https://codeberg.org/boud/maneage_dev/src/branch/maneage/README-hacking.md
_______________________________________________________
Reply to this item at:
<https://savannah.nongnu.org/task/?16228>
_______________________________________________
Message sent via Savannah
https://savannah.nongnu.org/
- [task #16228] Guix-Maneage synergy,
Boud Roukema <=