sts10 / bottle Goto Github PK

I think this is how the tar tool works, but need to do more testing.

Even if that is how tar works, it's a separate question whether Bottle should follow suit. I think of Bottle as an archiving tool, a context in which an accidental overwrite could mean lost memories.

If Bottle fails to find a age KEYFILE at the standard location of $HOME/.bottle/bottle_key.txt (via the if [ ! -f "$KEYFILE" ]; test), Bottle could check sdb devices looking for a /.bottle/bottle_key.txt until it finds one.

If two such attached devices have a /.bottle/bottle_key.txt file on them, Bottle could:
1.) compare their public keys to see if they're in fact the same key
2.) if they are not the same, ask the user which they would like to use.

The use-case here is that the user has their bottle_key safely stored on a USB stick.

Though the more secure workflow here is likely for the user to store their age key on a Yubikey and use an appropriate age plug-in...

Should it use a .bottle file extension, rather than the more open .tar.gz.age? Would look cooler, but would be more difficult for users to understand how to extract contents.

A Mac user reports:

❯ ./bottle.sh folder 

gives an error:

tar: --absolute-names: Cannot stat: No such file or directory
tar: Removing leading '/' from member names
tar: Error exit delayed from previous errors.

This sequence works as expected on my Ubuntu machine.

Wondering if this has to due with the --absolute-names flag, or whether that just happens to be in the error message.