I've had some time to live with starlark now, and a number of use cases that have cropped up where depending on a dataset with qri.load_dataset has been problematic. I initially thought this was a permissions problem, describing the issue as required "permission escalation" to allow access to datasets during the download special function.
While working through this RFC, I've come to believe the problem is the way we declare dataset dependencies. This RFC aims to clear that up while pushing the problem of access control down the road, but still make some progress on how access control will integrate with this RFC.
I've had some time to live with starlark now, and a number of use cases that have cropped up where depending on a dataset with
qri.load_dataset
has been problematic. I initially thought this was a permissions problem, describing the issue as required "permission escalation" to allow access to datasets during thedownload
special function.While working through this RFC, I've come to believe the problem is the way we declare dataset dependencies. This RFC aims to clear that up while pushing the problem of access control down the road, but still make some progress on how access control will integrate with this RFC.