[Re] A three-layer model for the flow of particulate suspensions driven by sedimentation

[IbrahHm]

Original article: https://doi.org/10.1063/5.0261889 (free preprint)

PDF URL: https://github.com/IbrahHm/replication_three_layer_model/blob/main/article.pdf
Metadata URL: https://github.com/IbrahHm/replication_three_layer_model/blob/main/metadata.yaml
Code URL: https://plmlab.math.cnrs.fr/ltoti92/sednash

Scientific domain: Mathematics Applied to Geosciences
Programming language: C++
Suggested editor: Konrad Hinsen

[khinsen]

@IbrahHm Thanks for your submission! We'll start looking for reviewers.

[rougier]

@IbrahHm In the meantime, if you can rewiew one of these submissions that would be much appreciated.

[theRizwan]

@khinsen I saw you are looking for reviewers on this one, and I would like to offer for part of it with a clear limit on what I can usefully judge.

I have just introduced myself on the reviewer application thread. I am an industry software engineer, not a geoscientist or an applied mathematician, so I could not assess whether the three-layer model is correctly reproduced or whether the physics is sound. I would not want to be the reviewer responsible for that.

What I can do carefully is the first two of the four acceptance criteria that the implementation can actually be built and run by someone who is not the author, and that the code and instructions are clear enough for a stranger to follow. Concretely, for this submission I would:

• build sednash from source on a clean machine and record exactly what was required beyond what the instructions state
• check whether dependencies, compiler versions and build flags are pinned or merely implied
• run whatever the article claims is runnable and report where the results are reproducible, where they are not, and where I could not tell
• note that the code is hosted on a CNRS GitLab instance rather than an archived deposit, which is worth a view on persistence

My background is build systems, packaging and reproducibility rather than numerical modelling. This year I fixed a defect in recast where a documented round-trip guarantee failed on unmodified input, verified with a differential harness over 5,076 generated inputs across six parsers, and a related fix in postcss-selector-parser. My C++ is reading-level rather than expert, which I think is sufficient for build and reproduction work but not for reviewing the numerics.

If it would help to have that half covered while you find a domain reviewer, I am glad to start now. If you would rather wait for a reviewer who can cover both, that is entirely reasonable and no reply is needed.

[khinsen]

@theRizwan Thanks for proposing a partial review - that would be very useful!

[theRizwan]

@khinsen Following up, since the label still reads NEED: Reviewer 2 and I did not want to leave the offer sitting.

Unless you would rather I wait, I will make a start this week and post my findings in this thread by 15 September: building sednash from source on a clean machine, recording exactly what was needed beyond what the instructions state, whether dependencies, compiler versions and build flags are pinned or merely implied, and how far the runnable claims hold up. Scope as before — the build-and-reproduce half only, not the numerics or the physics.

Two things worth settling first:

1. Is a written report in this thread the right format, or is there a form you would like me to use?
2. The code sits on a CNRS GitLab instance rather than an archived deposit. Do you want a view on persistence as part of the review, or is that an editorial matter for you?

If you would rather pair this with a domain reviewer before anything starts, say so and I will hold.

[khinsen]

@theRizwan No need to wait for a second reviewer to be named, you can do your review now.

A written report in this thread is exactly what we do at ReScience.

As for persistance, we strongly encourage authors to archive their repository on Software Heritage, independently of the forge they use. It's something I will check for myself!

[khinsen]

After reading the manuscript, I declare myself Reviewer 2!