what is the correct forking behavior with respect to files?

[hamilton]

Currently, forking does not copy files. That seems like perhaps the right approach, but I can see how this might cause confusion. At any rate, the notebook client needs to have some logic to clear the files part of the store when a notebook is forked.

[bcolloran]

i think copying files [*] is the correct approach -- for a user to be able to fork another user's work in order to experiment with it, they need access to the data associated with that notebook.

[*] this wouldn't have to be a literal file copy, but the new notebook needs to be able to access that same data, so some kind of pointer or symlink or literal copy is needed.

[hamilton]

I think this would be worth talking about more. It seems like the expected UX for forking would be literal copying, since the files may be cached data or files that are manipulated and changed / cached in the notebook, so the most conceptually simple approach would be to copy everything.

One other approach would be to give users the option of copying the files (if there are files)