The assistant archival walk-through describes how files get removed from local repo if moved into archive repos. What if the files in question aren't needed immediately (and can be archived) for one user of a shared repo, but needed by someone else on their own machine? git-annex directory arrangement propagation wouldn't work for this case, it seems. Could assistant handle this case?
It looks like having a browser of the directory tree in git-annex assistant webpage with say right-click menu with "get" and "drop" would be the most flexible.
Or maybe it could be more about integration with Linux etc file browsers (but then it's system specific, unlike with webapp).
I believe that the world has enough (or too many) file managers, and do not want to build another one just for git-annex. In any case "git annex get" or any GUI wrapping around it is not going to help with the described use case, because the problem is that the files are archived away in my offline drive, while you want to access them in your repository, which does not have access to that drive.
The way to handle this situation is to copy or move the files you want out of the archive directory, and back to the parent directory. This move of the file will be synced from your repository to all the other repositories. When the repository that does have access to the archive notices this, it will get the files that have been moved out of the archive. Normal assistant syncing will then arrange to get them transferred back to the repository of the user who had wanted them.
Of course, if the archive is on an offline drive, this won't happen until that drive is plugged back in.
But won't it also transfer the same files into repos of all other users (who didn't want the files)? On a related note, allowing moving files for everyone could jeopardise integrity of the master archive. Is there a way to limit what kind of ops the users can do (e.g. copy but not move)?
Re offline drives, the case I was thinking of is actually where each user has access to e.g. special remotes directly. So a manual get/put would solve the problem at hand.
Moving files shouldn't jepardize anything. If a user makes a change they shouldn't have made, it can be reverted.
If you have the archive drive available locally you can set your repository to manual mode and just use
git annex geton content you want. Or you could set it to be a backup repository and the assistant would get all content. Or write a custom preferred content expression for your repository to make it want files inarchive/stillusing/I think i agree that this should be implemented.
Not only to be able to "get/drop" from the assistant without the use of archive folders.
But mainly because i think this will be needed anyways, to do a "recover previous version of file" option. Which should definitely be in the web assistant.
We can't expect family and friends to figure out all command line stuff, thus they need to use the assistant. And i really think it would be a shame to not have an easy way to recover old versions, when the old versions are actually stored.
Always think about having a good WAF(Wife acceptance factor).
Just my 2 bits (and a haircut).