New RSE specialization: Open Source Developer

[Bisaloo]

Open source developer RSEs work on large pieces of software relevant for the wider research community. Oftentimes, it will be general purpose statistical or data analysis library, such as Stan or SciPy.

They are probably closer to industry roles than other RSE specializations currently in the paper but I still believe it makes sense to consider them as RSEs because:

• they need to understand the research lifecycle to inform the software lifecycle and their processes regarding breaking changes, new features rollout, etc.
• they need to be able to communicate efficiently with researchers who are the majority of their userbase and contribute most bug reports & feature requests
• it's more common for them to have a research background (or even to currently be affiliated to a research structure) than on average in industry

The name may sound similar to Open Science RSE but I believe they are very different: the Open Science RSE will mostly focus on making research artifacts open & FAIR, while the Open Source Developer RSE will play a role earlier in the research process, by providing low-level libraries used to produce these research artifacts. In some way, the closest specialization is probably the Web-Development RSE but they focus on very different product and use very different languages & frameworks.

[CaptainSifff]

Hi @Bisaloo , we discussed this last friday, and wre not quite sure which direction to take. I agree that any Open Source Developer in research is doing RSE work, but isn't it the other way around? In your domain, you are free/obliged to do OSS and hence you have the freedom to work on OSS and produce the respective OSS? We thought that it's a task of an RSE to consult in doing proper OS Development. What do you think?

[mschwarzmeier]

Hi, discussion seems a bit cooled down. Anyways... my take on this is the OpenScience-RSE, as currently described in the master version seems to be more of a domain RSE, that works in an open project. Therefore artifacts might be simulation data (and similar) and to make things reproducible a container or "reproducibility" with scripts. An OpenSource-Dev-RSE will have no (or little) simulation output but his/her artifacts will be the software as packages, like those packages Bisaloo gave. E.g. interoperability, speed, efficiency is much more important than for a domain RSE. Exaggerated: OSource-Developer-RSE run unit tests, OpenScience-RSEs run simulations. I see overlap with the HPC-RSE, but not so focused on hardware and administration, legal, environmental, scheduling and proposals.

[knarrff]

I think this is now covered with a combination of "open science RSE" and "project/community manager RSE".

[knarrff]

(don't hesitate to re-open - I am just in the process of cleaning up a little)