Some comments and observations

[jspaaks]

Some comments and observations based on https://github.com/sdruskat/software-authorship/blob/02fa9dc630c687ca0e2963359994b4f1d58721f3/index.Rmd

1. Name of the project is a bad idea marketing-wise -- I suggest getting rid of the special symbol
2. In the taxonomy, use US-en spelling, e.g. Conceptualization as opposed to Conceptualisation
3. In the taxonomy, use Data instead of Data curation, the latter is at odds with the high abstraction level that the rest of the roles have
4. "Who is an author" section: "the four criteria" I think you mean "three"?
5. Who is an author? section
   1. According to this section, Authorship is differentiated from Contributorship based the type of contribution (e.g. a Data curation role never warrants rising "above" the level of a mere contributor, regardless of how impactful that contribution may have been) and on the substantiveness of the contribution: "Those who do not meet one of the four [three] criteria due to the insubstantiality of their contribution should be acknowledged as contributors". So if I make an insubstantial contribution to, say, Conceptualization I should be credited as a contributor? I think what you want instead is that anybody who makes a substantive (maybe less ambiguously, "helpful") contribution is deserving of being credited somehow, and that one type of contribution is authoring the software.
   2. As a second point, I think there is something going wrong with the semantics of bullet 3 "safeguarding the [...] sustainability of the software project": I think this definition may include someone who, say, provides resources (hardware, money) to make sure the project lives, right? I find it odd that the recommendation says we should describe this role as "This person wrote the software".
   3. Maybe it makes more sense to start with the list of contributor roles, and then follow that up with how authorship is derived from that list (probably based on the contribution being of type Conceptualization OR Development (but not Supervision IMO) AND being substantive (more than just "helpful").
6. Post-it image, lower left corner: typo in "Principle investigator"
7. Section "What do you need to do as a project?" bullet 1 talks about "preferred citation", this is confusing because using preferred-citation from CFF goes against best practices and should be discouraged.
8. typo "authhorship" double h
9. FAQ 4 is non-committal, I think it's better to omit clauses like "it's different for each project" and "come up with a set of questions". I'd prefer it if this document would give me those questions :). As a user, of course I will adapt or ignore if I feel that's best -- no need to clause against that.

Hope you found some of this helpful!

[jspaaks]

One more thing I just experienced playing with crosswalks: if you choose to describe contributor roles at a high level of abstraction like it is now, that's easy for users because they don't need to learn the specific meaning of many terms. However, you'll find, as did I, that converting to essentially any other format becomes almost impossible. For example Conceptualization, does that translate to Allcontributors ideas? I think Conceptualization > ideas. Does it translate to CodeMeta 3.0's proposed Architecture, or Design? Again I think Conceptualization > Architecture and Conceptualization > Design. Suppose we choose CodeMeta 3.0's Design as the conversion target from your Conceptualization. Seems reasonable, right? Gotta choose something. But now you can get in a situation where a person's contribution would exactly fit CodeMeta's Architect role. When they try to write this down in your format, Conceptualization is the closest match, so that's what they choose. If that data is converted to CodeMeta at some point, you get the strange result that somebody who's clearly an Architect is now called a Designer.

So perhaps all of this means roles are only for human consumption? Or at least not for conversion? Existing systems would require updating before they could make use of the role metadata.