It does, thanks, Lev.  Its clear (-er.)  I see I need to reread the doc
(when Im not so swamped).  
John

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of
Novikov, Lev
Sent: Wednesday, March 09, 2011 8:53 AM
To: CICM Discussion List
Subject: Re: [cicm] Channel and Conduit

John,

On 2011-03-09 11:16, John Davidson wrote:
> Could you give me a few words on why the channel cannot be separated
> from the stream in a single domain.  The concern I had was that the
> control (is that the channel?) often needs to be isolated from the
> stream in several dimensions (e.g. integrity + secrecy.)  

Just to clarify, when I say "controller" and "stream" I'm referring to
the type of API objects that are returned to the client program; not
what is happening inside the crypto.

Also, I realize now that I need to qualify my statement more carefully.
In CICM there are three namespaces that produce "single domain"
channels:
CICM::Emit (Random, Pseudorandom, etc.), CICM::Answer (Hash, Sign,
Verify), 
and CICM::Coprocessor (Encrypt, Decrypt in a single domain).

CICM::Emit contains Controllers and Conduits because you may want to
emit
data out of the module to another domain (using a Controller; you
control 
the Channel, but don't see the data) but you may also want to retrieve 
the data into the current domain (using a Conduit; you control the
Channel 
AND you see the data).

CICM::Answer and CICM::Coprocesser only have Conduits because the client
program that creates a Channel in a single domain also sees the results
of that Channel (control the channel AND you see the data = Conduit).

Technically, this is a design choice-- we could have said that multiple
client programs are involved in a single domain operation (e.g., one to
create the Channel and another to send the data and receive the result).
However, coordinating this process for single domain Channels is harder
when there isn't something like an external port as there is in a
typical multi-domain high assurance scenario.

Regarding your point about integrity and secrecy:
CICM::Coprocessor channels come in several flavors including those that 
combine secrecy and integrity. Again, this discussion is about what
happens
in the client program, not what happens in the crypto.

See <http://tools.ietf.org/html/draft-lanz-cicm-cm-00#section-18>.

Let me know if this addresses your concern. If not, I can elaborate more
on the origin of the Controller and Stream concepts and why you can't
get
at them separately in a single domain.

Lev
_______________________________________________
cicm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/cicm
_______________________________________________
cicm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/cicm

Reply via email to