[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#869873: marked as done (Stretch regression: apt-key fails due to missing gnupg)



On Thu, Jul 27, 2017 at 03:09:50PM +0200, Harald Dunkel wrote:
> If gpg and its derivatives are so fragile, I wonder if there
> is an alternative?

gpg isn't fragile. Fetching keys and stuff directly from the internet is
not much of a problem in its normal mode of operation as you still have
the web of trust – you don't magically trust a key just because it is in
the keyring. gpg will in fact complain about not trusting it if you or
trusted peers haven't signed it. That needs a user to actively maintain
it a bit by refreshing keys, assigning trust values and co through.

That is different from the situation apt is in: If a key is in the
keyring it has ultimate trust, so that keyring must be protected with
your life – but on the upside you don't need to manage it manually if
you don't want to as it will be managed for you.

Argueably the behaviour of gpg is better, but has one tiny problem: Have
you got a verify connection to the archive signing key via the web of
trust? The answer is very likely no, due to it being a rather technical
thing at the moment.

So what apt and basically any other software does is having some kind of
ultimate trust database. Your browser e.g. has a giant certificate
store. And they all hope that changes to that database are done with
great care.

In apt it was just decided that we aren't going to invent our own way of
public-key cryptography and implement it, but just reuse what already
exists in the form of gpg by bending it to our needs – with the upside
that one day we might 'easily' change to the web of trust if it becomes
feasible for our usecase.


Best regards

David Kalnischkies

Attachment: signature.asc
Description: PGP signature


Reply to: