On Fri, 22 Aug 2003, Dave Crocker wrote:
Harald,
HTA> and should make no assumptions on
HTA> whether the trust is rooted in Verisign or psg.com, or is brokered
through
HTA> a mechanism that doesn't need a single root for its trust mechanism.
HTA> If "centralized conference control" doesn't mean that anyone with the
HTA> required resources can be a center, I think we're on a very wrong path.
The style of protocol needed for a single, centralized control
mechanism, versus a distributed and cooperative control mechanism,
typically is quite different.
The obvious example would be centralized dns control -- as we now have
for each node in the tree -- versus efforts to create multiple
authorities for any given node, such as the root.
Did you have in mind a way to reconcile this?
Different protocols ... Harald's only stipulation is that the protocol
shouldn't prevent anyone from creating a centeralized conference control
facility.
The distinction I see between centralized control and distributed control
is with centralized control, I must have a way to establish trust with the
control facility which I must trust to authenticate all other participants
in a similar fashion. With a distributed mechanism, I must establish a
trust relationship directly with each participant.