On 4/11/18 1:35 AM, Eliot Lear wrote:

Hi Robert,

A few additional comments below:

On 10.04.18 16:22, Robert Sparks wrote:

On 4/10/18 5:43 AM, Eliot Lear wrote:

Hi Robert and thanks again for the review.  Please see below for responses.  These are my personal views.  The WG chairs / shepherds may have different opinions.

Thank you for the very quick response!

On 09.04.18 19:57, Robert Sparks wrote:
Reviewer: Robert Sparks
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 wait for direction from your
document shepherd or AD before posting a new version of the draft.

For more information, please see the FAQ at


Document: draft-ietf-opsawg-mud-20
Reviewer: Robert Sparks
Review Date: 2018-04-09
IETF LC End Date: 2017-11-07
IESG Telechat date: 2018-04-19

Summary: Almost ready but with minor issues to address before publication as a
standards track RFC

Minor issues:

Section 3.15 is confused, and I don't think you'll get the implementation you
intend with the MUST in the current language. direction-indicated is not a
flag. The text about dropping should talk about matching the direction that was

Having reread this section (and perhaps I am a bit too close to the text), perhaps it is a bit confused.  How about something along the following lines:

3.15.  direction-initiated

   When applied this matches the direction in which a TCP connection is
   initiated.  When direction initiated is "from-device", packets that are    transmitted in the direction of a thing MUST be dropped unless the thing    has first initiated a TCP connection.  This node may be implemented in    its simplest form by looking at naked SYN bits, but may also be implemented
   through more stateful mechanisms.

  [RFC6092] specifies IPv6 guidance best
   practices.  While that document is scoped specifically to IPv6, its
   contents are applicable for IPv4 as well.


The quoted issues below are from my early review of -08. I don't think they've
been addressed or responded to. Apologies if I missed a response.

The document proposes "reputation services". It needs more words about
whether those exist, and what scopes the architecture imagines (an
enterprise might have a different idea of a reputation service than a
residence). There is a notion of "decent web reputations" in the security
considerations section. Who determines that? The security considerations
section should talk about attacks against the reputation services.
I think there needs to be more discussion of the PKI used for signing MUD

While this section is admittedly a bit vague, we need some operational experience to develop the appropriate use of PKI as an anchor for reputations.  This having been said, if you have a specific proposal for text, I'd be interested in what you have to say.
Do you envision enterprises or manufacturers creating a new set of anchors of trust, or are you hoping to reuse the web PKI or something else? If you don't know and all of these are on the table, mentioning it would help implementers avoid assumptions that could hinder deployment.

Thanks for the guidance.   In fact I think we need to more clearly specify the signature in terms of the purpose.  The tools (openssl) support S/MIME signing easily enough for CMS.  I think this is what we want, and it implies that we are *not* using the webpki, but instead just getting an S/MIME certificate.  This implies that the signers will have to be "admitted" on first sight.  Would that work for you?

Also see below.

Consider discussing whether the stacks used by typical things will let
them add DHCP options (or include bits in the other protocols being
enabled). If it's well known (I can't say) that these stacks typically
_won't_ provide that functionality, then you should punch up the
discussion of the controllers mapping other identifiers to MUD URLs on
behalf of the thing.

We did indeed add some text about this, almost verbatim to what you have above (I think at your suggestion).  This can be found in the introduction toward the bottom of page 9.
I'm not seeing it in -20. Maybe its in your working copy but not issued yet, or 9 above isn't where you meant to point? I did a quick search for DHCP through the document and didn't spot the discussion. Apologies if I'm just missing it.

Actually it's in two places.  See where we say in Section 1.5:

   It is possible that there may be other means for a MUD URL to be
   learned by a network.  For instance, some devices may already be
   fielded or have very limited ability to communicate a MUD URL, and
   yet can be identified through some means, such as a serial number or
   a public key.  In these cases, manufacturers may be able to map those
   identifiers to particular MUD URLs (or even the files themselves).
   Similarly, there may be alternative resolution mechanisms available
   for situations where Internet connectivity is limited or does not
   exist.  Such mechanisms are not described in this memo, but are
   possible.  Implementors should allow for this sort of flexibility of
   how MUD URLs may be learned.

Elsewhere we say:

   If a MUD controller is not able to fetch a MUD URL, other means MAY
   be used to import MUD files and associated signature files.  So long
   as the signature of the file can be validated, the file can be used.
   In such environments, controllers SHOULD warn administrators when
   cache-validity expiry is approaching so that they may check for new

So I'm hoping between the two we have you covered ;-).
Yes, I'll let this point go. You have well covered discussing controllers being able to map things to MUD URLs. You still don't explicitly discuss whether existing libraries would let things provide the URL in DHCP - that's what I was scanning for. But again, I'll let it go.

More below.

You suggest the DHCP Client (which is a thing) SHOULD log or report
improper acknowledgments from servers. That's asking a bit much from
a thing. I suspect the requirement is unrealistic and should be removed
or rewritten to acknowledge that things typically won't do that.

I propose to delete that paragraph to match that we do not wish DHCP servers to modify their state.  This would address your concern (also see below).

The security and deployment considerations sections talk about what the
need for coordination if control over the domain name used in the URL
changes. It should talk more about what happens if the new administration
of the domain is not interested in facilitating a transition (consider
the case of a young company with a few thousand start-up-ish things out
there that loses a suit over its name). Please discuss whether or not
suddenly losing the MUD assisted network configuration is expected to
leave the devices effectively cut-off.

The example you are talking about is a subset of what happens if the file simply doesn't exist.  At worst, one is left in a situation no worse than we are today: that is, someone will have to manually decide policy, or apply a default policy. This particular text is the subject of separate piece of work that Thorsten Dahm and Steve Rich are considering, and it is also the subject of some ongoing research.  That is to say: we might be able to do better in the future when we have some operational experience.
OK. I suppose the detail I was hoping to see was some insight into recovery (how my hypothetical company above could start providing new MUD files that the existing devices would be convinced to use). I'll accept that is "future work".


Right now, you leave the DHCP server (when it's used) responsible for
clearing state in the MUD controller. Please discuss what happens when
those are distinct elements (as you have in the end of section 9.2) and
the DHCP server reboots. Perhaps it would make sense for the DHCP server
to hand the length of the lease it has granted to the MUD controller and
let the MUD controller clean up on its own?

Ok, two issues:

 1. There was some text that should have been removed, referring to
    the DHCP server returning the MUD-URL as part of state.  With
    everyone's permission, I propose to remove that text.  No
    additional DHCP server state should be required to implement
    this option.
 2. I agree that we could be a bit more descriptive about what would
    happen if a reboot of the DHCP server.  I also like your idea
    about mentioning the lease time to the MUD controller.  With
    that in mind, how about the following text:

With that in mind, I propose the following edit:

DHCP servers may implement MUD functionality themselves or they may
pass along appropriate information to a network management system or
MUD controller.  A DHCP server that does process the MUD URL MUST adhere
to the process specified in {{RFC2818}} and {{RFC5280}} to validate
the TLS certificate of the web server hosting the MUD file. Those
servers will retrieve the file, process it, create and install the
necessary configuration on the relevant network element. Servers
SHOULD monitor the gateway for state changes on a given interface.  A
DHCP server that does not provide MUD functionality and has forwarded
a MUD URL to a MUD controller MUST notify the MUD controller
of any corresponding change to the DHCP state of the client
(such as expiration or explicit release of a network address lease).

Should the DHCP server fail, in the case when it implements the MUD
controller functionality, any backup mechanisms SHOULD include the MUD
state, and the server SHOULD resolve the status of clients upon its
restart, similar to what it would do, absent MUD controller
functionality.  In the case where the DHCP server forwards information
to the MUD controller, the MUD controller will either make use of
redundant DHCP servers for information, or otherwise clear state based
on other network information, such as monitoring port status on a
switch via SNMP, Radius accounting, or similar mechanisms.

wfm - thanks

There is tension between the paragraph in the introduction that characterizes
the manufacturer as the entity that provides the mud file and the third bullet
in the intent list about speeding vulnerability resolution for devices that
are no longer supported by the manufacturer (you intend here that someone other
than the original manufacturer or integrator is providing a new mud file).

Perhaps this is best fixed by dropping the phrase "by the manufacturer"?

Instead of "A light bulb is intended to light a room.", I suggest
"A light bulb is intended to light a given space." (There are lightbulbs
meant for outdoor use, and others for only inside cabinets or appliances.)


Instead of "We make use of YANG because of the time and effort spent to develop
accurate..." I suggest "We make use of YANG because it provides accurate..."

At the end of section 1.7, instead of "For these reasons only a limited subset
YANG-based configuration is permitted in a MUD file.", I suggest "For these
reasons, the YANG-based configuration in a MUD file is limited to the YANG
modules defined in this document." or something similar.
s/defined/specified/ but otherwise ok.

Just one clarification- this text needed to be wordsmithed a little bit more.  Keep in mind that the ACL model is supported as discussed in the draft.

In section 3.6, the parenthesized clause does not belong in the sentence in
which it currently appears. I suggest it belongs somewhere in the previous

I've removed it and rewritten as follows:

The intent is for administrators to be able to see a
brief displayable description of the Thing.

In section 3.13 you talk about classes that are standardized. Where does
someone find out about standardized classes?

They're described in the document
Ah. maybe in 3.13 you could say "classes standardized by this document or future RFCs"?

Ahh!  ok.  I see the confusion.  I think an example may be appropriate here (we are literally getting a lot of operational experience with this day by day now).

Suppose you have a manufacturer X who has products P, D, and Q. They each want to talk to the same controller or group of controllers.  Use of "my-controller" may invoke a process on a MUD controller that requires that the administrator populate the class for each device P, D, and Q.  Whereas with use of a controller class, the administrator only has to populate the class once.

Similarly, the use of the word standardized naked like that is probably unhelpful.
Can I infer you plan to edit it out or dress it more?
One could imagine, for instance, Fairhair or some other consortium deciding to create standard classes.

What I propose is two changes to facilitate better understanding:

 1. To include the simple example described above.
 2. To add an optional "documentation"  element in the "mud" container
    that consists of a URL that points to documents for each class,
    when so used.



With this, I'm puzzled about the use of the word standardized at all. I think I'm hearing that you expect MUD controllers to know about some well-known classes by convention and that groups like fairhair or someone else might make a list of classes that MUD controllers might collectively decide to build in knowledge of. Am I getting closer to the right picture? (This is opposed to a set of classes that are created by a standards action and listed in a registry somewhere).


OPSAWG mailing list

Reply via email to