Secrecy and user trust

[Joel Rees]

I scanned this thread and was a bit disappointed. Lots of apparently  
competent people arguing from false assumptions.

None of you know me from Adam. If we met at a key-signing party, you  
would have no reason to trust this guy with an odd scare and a week's  
growth of whiskers on his face. Face to face is only useful if the  
face matches a familiar pattern.

Shoot, I know most of you better from the patterns in your posts than  
I would know you if we met face to face.

You could show me your passport and other ID (including the  
electronic ones), and I would still know more about you from the  
patterns of your posts on the list.

In the real world, if Jack Smack comes up to me and tells me my  
cousin is introducing him, I check with my cousin, and I still don't  
really trust him until I get to know him. Dorothy Developer who is an  
acquaintance of a friend of a developer known by a couple of guys I  
met at the last Linux conference I attended has an advantage in  
numbers, but I still don't trust her programming until I have run her  
stuff a couple of times. And checked my logs

Chain of trust is an illusion. So is a web of trust.

What I do have is a running OS and log files in my firewalls. And  
memories of domain names I have surfed by.

That the keys posted on the mirrors and the main site match is good  
evidence of a number of people cooperating. That is why I trust the  
keys enough to try the OS, and I could trust them no more were they  
to be signed by someone chained back to verisign or whoever.

Speaking of verisign, trusting a single entity as a "root of trust"  
is not far removed from trusting some charismatic religious or  
philosophical type to be your moral root. The CAs can serve a  
purpose, but they tend to be used against that purpose more than for.

The best use of the keys is as a checksum. Isn't that clear?

Joel Rees