Trust

[Francis McCabe <frankmccabe@mac.com>]

I am afraid that I caused something of an 'upset' at today's telcon  
concerning trust :)

Here are the issues as I see them:

1. I am unhappy substituting participant for agent. My reasoning is  
that automated systems need to be trusted at least as much as people.  
There was no clear idea on the call for what participant should be  
replaced by; although removing the qualifier at the beginning about  
mostly human participants may be removed. In normal English usage,  
participants are essentially human in character.

2. The relationship between credentials and identity and action is  
*not* one of authorization. I am not sure at the moment what it should  
be but authorization is about the empowerment to perform actions, not  
whether the person doing them should be trusted. (The CIC has the  
power to take us to war; but I would not trust him if he asked us to.)

3. Some of the modeling around trust is inextricably linked to the  
issues covered in Section 3, and so I think that some of this needs to  
be moved there. That is because the appropriate level of abstraction  
for trust requires discussing the relationships between organizations,  
actions, participants etc.

4. A Section 5 discussion on trust should focus on what is needed to  
support clarity on trust and what is involved in running such a system  
(In fact, I see elements around trust in all three sections: 3,4 and 5.)

So, thinking more generally about this, I feel that (a) we are getting  
close to the total amount of content for the document but that (b) we  
are likely to need some reorganization. This is not reason to stop our  
first public review though.

Frank