This is a bit of a nasty post, growing out of 2 months of experience with this feature. First I'll explain why I think it is useless in its current form, and then I'll come to why I think it may be a fundamentally flawed idea.
It is useless, because it does not work.
While that may be a blanket statement, it is essentially true.
Ok, it does work, it even works with a PGP keyserver. The problem is that you can't know that it does, or if it does.
I found that it did work with an external email address, but not in a loopback test. It did sign all emails though.
But it does not log that anywhere, so if you have an important message that must be encrypted, you can't use it, because not only you can't know if it will encrypt it, but you can't even find out afterwards!
It does not work with keys that have multiple email addresses associated (PGP). But you can't upload the same key multiple times with single email addresses.
Now I've explained why the implementation does not work, I'll come to why I think it's a fundamentally flawed concept.
The whole point of encryption and signing is that you establish a trusted communication channel between sender and receiver.
But not only does the astaro save those keys without keyphrases and the private keys are accessible to administrators, it doesn't even verify the identity of the sender.
Anybody with access to subnets that are trusted by the astaro or with an account can use every key that is present in the system by changing his sender address.
The second obstacle is a problem of the implementation, making it useless. The first one, that the keys are not stored securely inside the system, can't be surmounted, I think.
So the whole system does not work, and shouldn't be used in any circumstance, because it lets the users (and the receivers of the signed mails) think that there is some added security which in reality does not exist.
This thread was automatically locked due to age.
