How to archive software, and how important is this?

[mfenner]

• is citing software from a website out of scope?
• is citing software from a code repository out of scope?

[moranegg]

Archiving software is important for software preservation purposes but also for linking software with a consistent Knowledge Base.
Concerning the questions: if citing software is only possible via software papers it can be limited.
When producing datasets from non-scientific software available on a website or a code repository, is that out of scope?

[mfenner]

@moranegg these are questions that came up in our software citation WG call today. I personally think that software papers are a special case and should not be the only way software can be cited.

[moranegg]

@mfenner sorry I couldn't be present at the call today.

[zoidy]

Archiving is important for the reasons @moranegg mentioned. I'd add to it that research institutions may want to hold on to their IP via their institutional data repositories.

In terms of the "how" part of the question, #16 and #5 are closely related issues that need to be addressed simultaneously. E.g., for versioning in the context of the Software Citation Principles, questions to address are

• should versions be archived under separate DOI's (for specificity), or a single DOI (to avoid the dilution of credit).
• Although a hybrid approach (single DOI for the concept of the software, separate for versions) is a potential solution, repository systems don't generally support this. How to engage providers to add this support? What to recommend in the mean time?
• From the researcher perspective, having multiple identifiers point to variants of the same thing might be detrimental to encouraging software citations ( e.g., "why do I need to citations to essentially the same thing?", or copying and pasting others citations may result in a citation to the incorrect version for reproducibility purposes)

[Melissa37]

We have a question (eLife) - should we be encouraging authors (or doing it for them) to publish their code on Zenodo so that we have a stake in the ground and a DOI?
Currently we fork their code in Github at the time of acceptance of the research article and cite the main repo (which can be updated) in the reference list (with the commit as the version), but mention the eLife forked version in the text.

[danielskatz]

The software citation principles say that indeed, the specific version of the code that was used should be archived, and zenodo is a way to go this. However, we also said that this code is what should be cited, not the generic repo.

I would prefer to see what you are doing turned around - citing the specific version that is archived, and referring to the repo itself in-line or via an additional citation.

How you implement this is really up to you - what you are now doing seems ok, though I think it would probably be better for the authors to do this, to own their archived code, and to get into the practice of archiving it.

Also, I think that once Software Heritage becomes more used and gets some more tools around it, it will likely be able to fill some of this function of pointing to a specific archived version in the context of other versions and the full repository. I discuss this a bit on https://danielskatzblog.wordpress.com/2017/09/25/software-heritage-and-repository-metadata-a-software-citation-solution/