Re: [Gen-art] Genart last call review of draft-wu-l3sm-rfc8049bis-07

["tom p." <[e-mail]>]

< see inline>
Tom Petch


----- Original Message -----
From: "Qin Wu" <[e-mail]>
Sent: Thursday, October 26, 2017 8:02 AM

> -----邮件原件-----
> 发件人: Jari Arkko [mailto:[e-mail]]
> 发送时间: 2017年10月26日 4:53
>
> Reviewer: Jari Arkko
> Review result: Ready with Issues
>
> I am the assigned Gen-ART reviewer for this draft. The General Area
Review Team (Gen-ART) reviews all IETF documents being processed by the
IESG for the IETF Chair.  Please treat these comments just like any
other last call comments.
>
> For more information, please see the FAQ at
>
> <https://trac.ietf.org/trac/gen/wiki/GenArtfaq>.
>
> Document: draft-wu-l3sm-rfc8049bis-??
> Reviewer: Jari Arkko
> Review Date: 2017-10-25
> IETF LC End Date: 2017-10-11
> IESG Telechat date: 2017-10-26
>
> Summary: I'm not an expert on YANG *at all*. And not an expert on the
topic in question either. And I had far too little time to spend on this
long document.
> But as far as the textual content of the document goes, it seems
reasonable. I have a difficulty in assessing how complete and
implementable this model is however. Are there implementations?
>
> [Qin]: Yes, there are implementations, something broken in RFC8049
needs to get right.
>
> I did enjoy the classification of Internet connectivity as a special
case of cloud service :-) You may be onto something.
>
> [Qin]:Yes, internet connectivity is special example of public cloud,
in my understanding,:-)
>
> I did observe a couple of question marks or issues that probably
deserve some thought or small revisions.
>
> Major issues: -
>
> Minor issues:
>
> I'm not sure I fully understand the need for "SP MUST honour
<requirement>"
> language in the document. Are there parts of the described model that
they SP is *not* required to honour? Other than the explicit strict
true/false settings? And in any case, sizeable networks are likely to
have issues that might require negotiation/human involvement.
>
> [Qin]: Explicit settings are covered by sub-section of section 6.6,
such as device constraint, Site Location constraint/parameter, access
type.
> Unlike RFC8049, RFC8049bis change "SP SHOULD honor" into "SP MUST
honor".
>
> I don't understand how 6.9.1 can say there is no authentication
support but then 6.9.2 (encryption) talks about authentication keys. I'd
suggest some rethinking or at least clarification might be needed here.
>
> [Qin]: Authentication provides you access to a resource like a
computer or a network. Encryption provides confidentiality and controls
whether an object can be read or not.

Qin

I do not see that usage as common in the IETF and I think that it
reflects a fundamental flaw in this model such that the I-D should not
go forward in its present form, except that the same flaw is present in
RFC8049:-(

Authentication is the process of establishing that a person or object is
who or what they claim to be, using e.g. pre-shared keys or
certificates, in an authentication protocol.

Authorisation is the process of granting rights, privileges, access and
such like to an authenticated identity.

Authorisation without authentication is meaningless, a fantasy of
security

Scanning the I-D, I think that the usage of authorisation and
authentication is mostly correct.  As the I-D points it, there is no
support in the standard model for authentication.  Where the I-D is
wrong, perhaps dangerously so, is in the Security Considerations where
it says that because NETCONF may have used TLS or SSH to establish the
connection, and the NETCONF Access Control may be used to control
authorisation, then somehow this is secure.

You haven't a clue what identity has been authenticated and whether or
not they should be authorised to join a VPN.  This is a common problem
with applications running over a secure transport, the transport may
have autheticated an identity but how do you get that identity up to the
application for the application to verify that the identity really is
ok?  Channel Binding is one answer but rarely used.

Tom Petch

> Both Authentication and Encryption are site level parameters and
applicable to site connection. Pre-share key as encryption parameter can
be used, e.g., for routing protocol authentication.
> But pre-share key parameter is not authentication parameter, one
example I have for authentication parameter is authentication algorithm
which is not specified in the model,
> We will see how to clarify this in the model. thanks.
>
> In the security considerations, I would note that if these models are
used not merely for creation of networks, but also their modification,
the consequences of inadvertent or malicious modifications can severe
and network wide. Perhaps that could be discussed.
>
> [Qin]:Security consideration section has already considered both
creation of networks and but also their modification, see the text in
security consideration section as follows:
> "
>    The data nodes defined in the "ietf-l3vpn-svc" YANG module MUST be
>    carefully created, read, updated, or deleted as appropriate.
> "
>
> Nits/editorial comments:
>
> Section 6.12.2. s/fragmented/fragment it/
>
> [Qin]: Will fix this, thanks.
>