key question

David Shaw dshaw at
Sat Feb 27 21:02:54 CET 2010

On Feb 27, 2010, at 11:22 AM, Robert J. Hansen wrote:

> On 2/27/10 9:58 AM, David Shaw wrote:
>> Do you really mean to suggest that a US authority getting email 
>> headers - even without a warrant - is easier than typing a name into 
>> a search box on a keyserver?
> No.  You're right, that's clearly easier.  However, that only tells you
> whether someone has the technical capability to use encryption -- much
> the same way that a shotgun in my closet tells you I have the technical
> capability to commit murder.

Much as the email headers do in your example.  If the mail is not encrypted, the headers just show that it might be.  In practice, headers won't show much as the majority of modern mail programs have the capability for encryption of one sort or another, even without add-ons.  It's rarely exercised, of course.

> As a result, the possibility of law-enforcement officers checking the
> keyserver network doesn't seem to be a strong argument against the use
> of the keyserver network.
> The major exception is if you live in a jurisdiction where possession of
> crypto is itself a criminal offense.  If you live in Cuba and you're
> using GnuPG, then you should not have your key on the servers and you
> have a perfectly reasonable fear about people uploading your key there.
>> In any event, Rob, could you do me a huge favor and clarify what 
>> statement you are trying to make here?  Jumping into a mail thread 
>> late is always fraught with misunderstanding, but, I've re-skimmed 
>> the thread, and I'm honestly still not sure what you're trying to 
>> say.
> His position seems to have shifted.  At some points he's said,
> "What's not to agree with in my statement that not everybody wants to
> put their keys on the keyservers?"
> I fully agree with this.  However, he also seems to be advocating the
> advice of "generally speaking, it's a good idea to put keys on the
> keyservers" be changed to "generally speaking, it's not a good idea to
> share public keys without the key owner's explicit permission."
> This is a pretty big change in the conventional wisdom.  Before I'll
> sign on to that I'll have to see some strong reasoning, and I haven't.

I agree that "generally speaking, it's a good idea to put keys on the keyservers".  I don't know if that makes it conventional wisdom, or who the arbiter of such wisdom might be, but clearly a very common use of OpenPGP is for encrypted mail.  If you want encrypted mail, putting your key on a keyserver is very helpful in reaching that goal. The word "generally" takes care of the exceptions (as there always exceptions for one reason or another).  So basically, yes, if you're using OpenPGP, keyservers are great.

With regards to the second statement, you give a great reason yourself a few paragraphs up: "If you live in Cuba and you're using GnuPG, then you should not have your key on the servers and you have a perfectly reasonable fear about people uploading your key there".  Is that not a good reason to request that a key stay off the keyservers?  I don't find the behavior *behind* this reason very good, as if someone lived in a place where encryption was banned, they'd be foolish and naive to think that their key would stay off the keyservers merely because they requested it - one accident, and it's published, and no way to withdraw it.  People who live in places where encryption is illegal need to do a lot more than simply not send their keys to a keyserver if they want to remain safe.

Personally, I don't find most don't-publish arguments (spam, traffic analysis, etc) compelling, and I correspondingly do send my key to the keyservers (in my case, it would be particularly silly not to).  However, I never send anything to the keyservers (or publish otherwise) if it isn't mine.  I don't know what their situation is, and it's not up to me to decide it for them.  Even if I did know their situation, as in the Cuba example above, and disagreed with them on how to handle their key, it still is not my key, and not my decision to make. I don't know if that makes it conventional wisdom, but I have acted that way since I became involved in the OpenPGP world many years ago.  Whether it's wise or not, I'd at least hope it's common politeness.

Keys ending up on keyservers contrary to the desires of the key owner has been a problem for a long time.  Note the addition of the no-modify flag when OpenPGP was first published as an RFC in 1998.  That was added after experience with PGP 2.  The whole point of that flag is to only allow the owner to publish their key.  Similarly, note that the PGP Global Directory only allows key uploads from the key owner, avoiding this problem.  The earlier PGP "certserver" had the capability, though I don't believe it was always turned on.  Clearly this is enough of a problem that work was done to avoid it.

> For myself, I do not send keys up to servers without first checking it
> with the recipient.  This seems like good manners to me.  However, I
> don't view it as mandatory and I don't think we should view it as the
> appalling breach of morality that MFPA seems to.

So you are saying "I do not do this".  And MFPA is saying "I think nobody should do this" ?

Where's the problem?  From where I'm standing, that looks like something that passes for rough agreement, or - dare I say it - conventional wisdom. ;)


More information about the Gnupg-users mailing list