RE: Question on security document

[Robert Snively <rsnively@Brocade.COM>]

Folks,

 

I agree that this is in many respects an editorial issue, but it is one with profound capability

for creating severe technical discrepancies which will delay the creation of interoperable

implementations.  We must not only resolve the technical content (which we are near

to doing), but document it in such a manner that people will build it only one way,

the way we want.

 

Bob 

 

 -----Original Message-----
From: Charles Monia [mailto:cmonia@NishanSystems.com]
Sent: Thursday, October 18, 2001 5:22 PM

Hi Folks:

 

I hope we do not delay closure on the technical content in order to correct what seem to be editorial issues.

 

Charles 

-----Original Message-----
From: Robert Snively [mailto:rsnively@Brocade.COM]
Sent: Thursday, October 18, 2001 4:08 PM

Franco,

 

The problem is that the format of the document does not differentiate between

the tutorial information and the language that is to go into the Standard RFCs.

 

I would expect that we should have a formal clause for each document saying

something to the effect:

 

"and because of the knowledge carried in the above tutorial, the [FCIP, iSCSI, iFCP]

document should contain the following text to properly describe the expected

behavior.  For the actual standard requirements, see [FCIP, iSCSI, iFCP]."

 

This clearly separates the result of the discussion (text to be shoved into a

Standard RFC which has no authority in this document, but will have authority when it is

shoved into the Standard RFC, perhaps with editorial or technical modifications)

and the tutorial discussion itself (which explains

without any capitalized requirements why it is nice to do each of the things

proposed in the text to be included in the Standard RFC).

 

Bob

-----Original Message-----
From: Franco Travostino [mailto:travos@nortelnetworks.com]
Sent: Thursday, October 18, 2001 10:06 AM

We have routinely taken our text into the security draft, and we had it reviewed and scrubbed off there by a good group of security experts. We routinely took selected things back into the authoritative document (iFCP in our case, but it could have been any one the three protocols). The use of the very same textual conventions is key to this high-bandwidth interaction, as well as to keeping the documents in sync as we repeat the cycle over and over.

The security draft also serves the purpose of keeper of the knowledge, hence it does not fit the normative text  role.  Rationale text and data (e.g., the speed/cycles figures) will hopefully be a good companion text to many Standard RFC's readers. As well as new RFC writers (good data outlives bad theory :-).

I don't see a conflict between MUST words and Informational RFC, the latter clearly and unambiguously dominates over the former.

-franco

At 11:35 AM 10/18/2001, Robert Snively wrote:

Paul,

I rather prefer to see the security mandates carried in
the primary documents, as our IETF mentors originally proposed.
That way, there is less possibility of inconsistent
language that may change the primary documents.  Then the
security document would be a tutorial and remain informational.
This will simplify the standardization process, because
each of us will only have to read the security document for
guidelines to our development of the security section of the
primary documents.  If we were to make the security document
the standard, then each of us will have to
read the entire document to make sure that some global
restriction does not fall unintentionally on a particular
protocol. 

If we do choose to make the security document the authoritative
document (which I would vote against), then it needs a major
rewrite to clarify, separate, and make more precise the requirements for
each protocol.

Charles, 

If the security document was intended as a draft of the security
sections for each protocol, it needs a major rewrite to separate
the particular language to be dropped into each document from the
tutorial information.  It then needs to formulate much more
clearly and formally the language that will be dropped into the
primary documents.  That would be okay with me, but the draft
should indicate that the language to be dropped into the
primary documents is only proposed and that the actual applicable
standards information is contained in the primary document.

Bob

> -----Original Message-----
> From: Paul Koning [mailto:ni1d@arrl.net]
> Sent: Thursday, October 18, 2001 7:37 AM
>
> Excerpt of message (sent 17 October 2001) by Charles Monia:
> > Hi:
> >
> > I had assumed that one goal of the document was to set
> forth the language
> > necessary for each spec to pass muster with the IESG in the realm of
> > security.  If that's correct, I'm concerned that the
> suggested change may
> > compromise that intent.
> >
> > Charles
> > > -----Original Message-----
> > > From: Robert Snively [mailto:rsnively@Brocade.COM]
> > > Sent: Wednesday, October 17, 2001 4:31 PM
> > > To: 'ips@ece.cmu.edu'
> > > Subject: Question on security document
> > >
> > >
> > > I have a recommendation to the authors of the security
> > > document:
> > > ...
>
> The problem, as Bob is right to point out, is that informational RFCs
> by definition cannot establish requirements on implementations.  So if
> you want there to be security requirements that apply to iFCP, iSCSI,
> and so on, they can only be stated in standards track RFCs, not
> informational RFCs.  It may be a separate document or part of the IPS
> document it applies to.
>
> So one answer is for the security draft not to be informational
> anymore.
>
>       paul