Name of the repository

[giovannipizzi]

A discussion has started on an email thread, that I report here, allowing for further discussion (I think there is no confidential content so I just post it here).

From @ConradJohnston

I preface this by saying I think 'common' is okay , although this may be just because I've become used to hearing it.
I would urge that a "hallway usability test" be done with whatever the working name is to gauge how intuitive it is before committing to it forevermore.
a. "Common Generic workflows for the computation of materials properties using different quantum engines"
b. "Common Universal workflows for the computation of materials properties using different quantum engines"
Generic as in not specific. It has the same issue that 'common' has in English that it could imply a 'cheapness', although in scientific circles, hopefully this is not what first comes to mind.
Alternatively, 'universal' as in "universal serial bus" - unlike the 'common' or 'generic', universal only really has one obvious meaning in this context. The gotcha here is that it's really the interface that we're describing.

From @greschd

I think "unified" was another contender. All of them sound right to me when prefacing "interface", with "workflows" it's a bit less obvious.
Both "generic" and "universal" carry, to me at least, the connotation of "can do anything".
Ultimately I'll let you all who worked much more on this decide the naming, at the end of the day any of the choices work fine.

From @vdikan

Personally I would go with generic workflows, in analogy to generic methods in OOP.
It might not be linguistically correct but I guess it already is a common tongue in the
software world to designate a common interface to be implemented for a number of classes
(say, our codes) and later called by their instances (say, our calculations).
In a sense, not quite "can do anything" but rather "can be performed by anyone who is told, how to".

[giovannipizzi]

Personally I like the idea of calling this project the "Unified workflow interface(s)" (along the lines suggested by @greschd).
So, e.g., calling the repo unified-workflow-interfaces and using as a title "Unified workflow interfaces for the computation of materials properties using different quantum engines"

This to me seems very powerful.
The issues I have with the other options are:

• universal: for the way I interpret it, not only it seems it can do everything (a "universal tool", like a swiss-army knife), but even more critically it might pass the message that it represents the only way to do it. I don't want however to pass such message (we're still limiting, in practice, to AiiDA workflows, and we made specific decisions on how the "common/unified" interface should look like). The same goal (e.g. relaxing structure) could be achieved with a different common interface (e.g. 'material' instead of 'structure', different set of inputs, ...) and also with other workflow engines. [Note: I don't think we need to put AiiDA in the name of the repo, though, since it's already in the organisation: aiidateam/unified-workflow-interfaces]
• generic: The first meaning that comes to my mind is not the one that @vdikan suggests but that of "something that can achieve multiple goals, but to achieve this compromise it might not be optimal": like "it's a generic tool, not one tailored to solve a given goal in an optimal way", which (to me, at least) has a bit of negative connotation for our goal
• common: while it might still carry a bit of the same connotation of "generic", I get more quickly the meaning of "the same interface". Still, I personally like much more "unified workflow interface" (but would be ok, as a fallback, to e.g. rename to common-workflow-interfaces; I think speaking of common workflow interfaces rather than common workflows is more accurate).

[muhrin]

Just to chime in on the name debate. Personally I'm fine with 'Common...' and ultimately am happy to go along with whichever choice the people who spent most time/effort on this work decide.

There is one technical point I want to raise: These workflows are not 'generic classes' in the technical sense because generic classes (also called templates) [1] are specifically classes where the algorithm is pre-specified but the types are not (for example a list in C++ where the type is specified at compile time but all the methods are already present in the template).

Interfaces is more technically correct (assuming there are no implemented methods in the base classes in which case they are 'abstract classes').

Overall, I think I also like 'unified' the most.

[1] https://en.wikipedia.org/wiki/Generic_programming

[nicolamarzari]

Dear all,

I have to say that also to me generic brings a sense of compromise. The plus for common is that it make a nice contrast to different (common wokflows/different engines) and it also echoes community, or common good, that is what we are doing here - we are having shared (community) workflows for the common good.

[sphuber]

As already mentioned by others, the real ambiguity of the word "common", or any of the other suggestions, is when the word "interface" is omitted, because the adjective really applies to the interface and not the workflow itself. That being said, I think I also like "unified" more than "common" since the it carries a stronger positive connotation, so I would be in favor of that name change.

Regarding the title: in the paper I have tried to stress how having a unified (optionally transparent) interface drastically increases the reusability of a workflow, so I think it would be great if we can get that term in the title. I was thinking of something like:

Reusable workflows through unified interfaces for the computation of material properties using different quantum engines"

This has the added benefit that now the adjective "unified" directly applies to the interface as it should and so there is no longer a problem of ambiguity. In the text we can explicitly define that when we use "unified workflows" we are using it as a shorthand for "unified workflow interfaces".

Regarding @giovannipizzi statement

Note: I don't think we need to put AiiDA in the name of the repo, though, since it's already in the organisation: aiidateam/unified-workflow-interfaces

I am not sure I agree here. The name of the repository is not always going to be mentioned in combination with the Github organization, and the interface (for now) can only really be implemented using AiiDA. So I think making it aiida-unified-workflow-interfaces is better. The additional six characters should not really be a problem I think.

[nicolamarzari]

I think that the title suggested by Sebastiaan ("Reusable workflows through unified interfaces for the computation of material properties using different quantum engines") is very precise, and says everything that we want to say. On the other hand, conciseness is also very important - a short title is always more impactful. So. I would tend to highlight these key concepts in the abstract, starting even with the first sentence (e.g. We propose and develop reusable workflows though unified interfaces...) but I would keep a shorter title such as "Common workflows for materials properties using different quantum engines")

[espenfl]

It is nice to see discussions on this topic. I am also with the others and in favour of unified over common. Also, as mentioned before, without something like interface it is simply not describing what is in this particular repository. When it comes to the manuscript one could argue that this also describe the results of an execution of the workflows for each respective plugin, so that would of course be more in the grey zone with respect to the requirement of interfaces. But maybe better to err on the restrictive side.