Closed aschetti closed 3 years ago
@aschetti looking into this now, and I remember why I didn't follow up on this earlier: Because users need to make a PAT, which is a substantial bottleneck (cf: the GitHub authorization is currently the largest bottleneck for users). https://docs.ropensci.org/osfr/articles/auth
Do you have any thoughts on this?
Yes, I can imagine that a PAT could be a bottleneck for inexperienced users. I can't think of a way to automate the token creation on the OSF but, once it's available, perhaps worcs
could automatically authenticate users at the beginning of each session or, better, generate a .Renviron
file in the project directory that defines OSF_PAT
. During the first initialization of the worcs
project, the user could be asked whether they want to connect to the OSF (yes/no), whether they already have a token (yes/no, with a link to instructions on how to do it), and paste the token. Having a dedicated function could also allow users to connect to the OSF at a later point in time, only if necessary (e.g., because data are too big for GitHub).
This is a really good idea, thank you for brainstorming it!
@aschetti I've been investigating this, but now leaning towards not supporting it.
Reasons:
1) My students found some non-reproducible code, which failed because functions in osfr
were not deprecated properly
2) Our workflow recommends hosting the files on GitHub - not on OSF storage. This is because GitHub offers the advantages of version control, easy updates from the console, online editability, and collaboration. The main function of osfr
is currently to interface with OSF storage, so if we're not using that - the package is not very useful
3) osfr
does not allow you to add preprints to an OSF project, currently.
@cjvanlissa I completely understand! After all, osfr
can still be used in combination with worcs
, you don't necessarily need to facilitate its usage if it doesn't fit well with the philosophy behind your package. Thanks anyway for the nice brainstorming!
During the awesome WORCShop at the Open Science Community Rotterdam, I noted that the preregistration created within
worcs
is pushed to GitHub but not to the Open Science Framework. It may be a good idea to allow users to optionally push there as well: researchers in several social science disciplines are actively using the OSF Registries as preferred prereg service, because it can be connected to a companion OSF project with data, code, and materials.One possibility could be to push the preregistration created within
worcs
to a specific OSF component (e.g., called preregistration) within a project, which would then be linked to the OSF Registries. The R package osfr already offers a nice entry-point to the OSF, so perhaps aworcs
-osfr
integration could be useful.Once the connection to the OSF is established the user can decide whether to push only the preregistration or also something else, e.g., data that are too big for GitHub.
What do you think?