Re: [Isms] wg last call followup - sshtm

["David Harrington" <ietfdbh at comcast.net>]

Hi Tom,

inline. 


> > I was totally confused by this whole thing myself. Then, to answer
a
> > question from Joe, I researched it fairly thorughly through the
> > RFC341x series and realized my whole understanding was 
> wrong, and then
> > explained it to Joe, who finally understood it. Now Tom has 
> a problem.
> > (And I still slip into my old understanding and have to 
> remember this
> > new approach). I expect that most operators will have an 
> understanding
> > similar to my original understanding, which I think was the
original
> > intention of the SNMPv3 WG. I think we are turning things on their
> > head here, and we better be sure operators understand.
> >
[...]
> > alice is actually the sender, not the recipient. That one
obviously
> > slipped by.
> 
> David,
> 
> Actually, I see this one as ok:-)  It tells me that there is 
> a recipient and
> that the identifier for that recipient has two values, one, 
> alice is used in the
> NO, the other, rtr-nyc4, is used in the NR.  

And that is the confusion I had (or have) as well. I think saying that
alice is the recipient is wrong, if the bob@ format is used. 

In USM, the sender and recipient are the same.
In SSHTM, the sender and recipient are the same, **if** a user@
address is not used.
In SSHTM, if a bob at examp;e.com:PPP address is used, then alice is the
sender and "bob" is the recipient. 


> Well, ok as in 
> correct, less ok as
> in is it easy to read and understand.  I mentally duplicate 
> that text replacing
> recipient with originator and that is also ok ie I import the 
> Architecture/USM
> thinking that a transaction has one principal only now we use 
> two different
> identifiers for that principal, one in NO, the other in NR.
> 
> Whereas, as I say in my e-mail to Juergen, I cannot get the 
> right meaning out of
> the e-mail he posted yesterday.
> 
> "The point is that for notifications, access control
> uses securityName to identify the notification receiver while SSH
uses
> the SSH user name to identify the notification sender. In many
cases,
> these will not be the same. So the user at host format allows to make
> this important distinction. Is this understandable?"
> 
> That does seem back to front to me but perhaps there is a 
> reading of that that I
> am missing.

I believe that for notifications, SNMP uses securityname for access
control for the message **sender** not the notification receiver.

Because in USM the sender and receiver securityName/principal is the
same, this concept of access control for the sender, not the
recipient, did not come easy. And if Juergen said "access control uses
securityName to identify the notification receiver", then he is also
working under the same bad assumptions I was working under before I
grok'd the new reality.

> 
> Tom Petch
> 
> > I think it is unrealistic to expect that we will have this 
> all sorted
> > out, and find all the incorrect usages, by the end of this 
> week. I do
> > not expect to have another revision by deadline. If Wes or Joe are
> > available and in a masochistic mood, maybe they'll get another
> > revision done.
> >
> > The next revision we publish will have to explain this whole thing
> > better so we are all on the same page (see there is a 
> benefit to only
> > having 6 people in the WG; imagine the difficulty if more 
> had to be on
> > the same page). Text welcome. Then we can really start to
> > fine-tooth-comb these documents for inconsistencies.
> >
> > dbh
> >
> <snip>
> 
>