We all have passwords. Lots of passwords. Be it for hosts or websites or disc-encryption. Noone can memorize a thousand different passwords for every account one needs. So we have two possibilities:

1. use a small number of default-passwords or
2. use a password-manager

Most people I know chose a combination of the two. But having shared passwords between different accounts makes you very vulnerable in case of targeted attacks (or even automated attacks, if the attackers tool is smart). So what you really want is different passwords for each and every account.

For that to work for daily use, you want an integrated wallet, like the KDE wallet, the Gnome keyring or the Firefox password manager. But for longtime archival to look up that password you had for that old mail-account somewhere, you want something seperate that can hold all kinds of different passwords and is more or less plattform independent.

For a long time I maintained my list of passwords with gpg: I had an encrypted file with all my accounts. To access one of the passwords, I could do a gpg -d passwords.txt.gpg | grep username. To add or change a password, I had to save the decrypted file on disk, edit the record, encrypt the cleartext again and then shred the remaining cleartext file. This was really cumbersome.

What I really wanted was a simple independant password-manager for the command line.

This is what pwsafe is.

It has a pretty simple cli. Every record has a group.name string as primary key and has username, password and optional notes as properties. The commands to access the records are straight forward:
pwsafe -l [REGEXP] — list passwords matching regexp
pwsafe -a group.name — add a new record
pwsafe -up [REGEXP] — print records matching regexp
pwsafe -e [REGEXP] — edit existing records matching regexp

The grouping system becomes really natural after some time. For every operation you have to enter a passphrase to decrypt the on-disc password database. The on-disc file format is compatible with the windows tool password-manager.

I’ve taken a quick view at the source-code. It tries to mimick the behaviour of Schneiers password-safe. That means it uses (openssl) Blowfish in CBC mode and tries to keep the passphrase in non-swappable memory (needs suid-root for that to work). The passphrase is directly fed into blowfish as the encryption key. That means if you supply an 8-character passphrase you only get like 40^8 bit of randomness. So you better chose a strong passphrase. To get at least a 64bit strong key from about 40 keys on the keyboard you need about 12 or 13 characters.

I didn’t look for any parser vulnerabilities, but so far it seems the author knows what he did and I can only suggest pwsafe :)