Allow using existing key file on login #688
Conversation
|
Seems ok, I'm a little skeptical on it's value, but I'm overall not against this |
|
(you can ignore the clippy failure, being fixed in #686) |
trygveaa
force-pushed
the
allow-using-existing-key-file-on-login
branch
from
678163b
to
54fd21b
Compare
I'd argue that the first commit is useful at least. After registering, I logged out and in again because I like to confirm that the login works and got the credentials I expect. On the key prompt, I assumed it would just use the existing key if I didn't specify anything, so I ended up with an empty key file. The second commit could be better. I think it makes sense to verify that the key is valid on login, but it would be better to actually verify that it's a valid key. Just checking that it's not empty isn't that useful, especially since you shouldn't end up with an empty key after the first commit. (And I see now that the sync panics if the key isn't valid, so there shouldn't be any security concern with an empty/weak key here). If we were to add a check that the key is valid, I think it would make sense to also do that check for key values specified at the prompt. |
There was a problem hiding this comment.
I'll also ask for @ellie's judgement, but I am happy to accept this
If the given key on login in empty, keep the existing key file rather than overwriting it with an empty file. This is useful if you log out and want to log in again and still use the same key, or if you have copied over the key file rather than providing it as input.
Before the previous commit, an empty key file would be created if key wasn't specified, and after the previous commit, the key file would not be created if the key wasn't specified and stay empty if it was empty. Now the log command checks the key file if a key is not specified and exits with an error message if either the key file couldn't be opened or is empty. If a key is specified, the key file is just created with it as before.
After reading the key either from an existing key file, or from the user input, validate that the provided key is valid (rather than just checking that it isn't empty). If no key file exists, create a new key instead of erroring out.
trygveaa
force-pushed
the
allow-using-existing-key-file-on-login
branch
from
54fd21b
to
a4f7540
Compare
I changed it to validate that the key is valid, both when read from an existing file and from user input. I also changed it to create a new key if the key file is missing, rather than erroring out. And I rebased on main to resolve the conflicts. |
If the given key on login in empty, keep the existing key file rather
than overwriting it with an empty file. This is useful if you log out
and want to log in again and still use the same key, or if you have
copied over the key file rather than providing it as input.
Before the first commit, an empty key file would be created if key
wasn't specified, and after that commit, the key file would not be
created if the key wasn't specified and stay empty if it was empty.
With the second commit the log command checks the key file if a key is
not specified and exits with an error message if either the key file
couldn't be opened or is empty. If a key is specified, the key file is
just created with it as before.
(Btw, great talk at fosdem!)