Enrollment over Secure Transport
draft-ietf-pkix-est-09

[Ballot discuss]
I'll end up as a yes for this but I have a one more thing
I'd like to DISCUSS first:

(1) cleared

(2) 3.2.1, .well-known URLs will often result in HTTP
re-direction. Don't you need to say more for HTTP
re-directions, e.g. that whatever is required for the
original server cert is also required to be true of the
location to which the HTTP client is re-directed? How
does the MUST for tls-unique values from the 1st TLS
session in 3.5 work with re-direction? I suspect that you
need to either work out how 3xx responses are done or
else explain it a bit more to me, because I didn't get
that it works now.

(3) cleared

(4) cleared

[Ballot comment]

-- this was a discuss but is now a non-blocking comment.
(I still don't like /.well-known/est/arbritaryLabel1 since
that ain't well-known!)

(1) I don't get how the client discovers the
"arbitraryLabel1" etc in 3.2.1? Don't you need to say
that they're assumed to be known to the client? (Or
provide a discovery mechanism, which sounds like too much
work.) I'm also unsure of the point of using .well-known
with those labels, but can live with it.

-- older comments below

- intro, why call out TLS1.1 and have no reference to
5246? I think at least adding 5246 would be good since we
would like folks to use TLS1.2 for this (and for
everything). I'm fine if this says that it doesn't really
matter (in functional terms) what version of TLS you use
which is what I think you're trying to say.

- intro, would it be worthwhile also saying that this
doesn't have any relationship with 6712 (which updates
4210) and sort-of attempts to do the CMP-like equivalent
of this? The only reason to add soemthing like that would
be to offset reader confusion if they go look at 4210,
see its updated by 6712 and then are puzzled that this is
very different from 6712. (And yes, all involved, incl.
me, are sorry about the 20-th century screw up that lead
to us having both CMP and CMC;-)

- 1.1: TA as the abbreviation for "Third Party Trust
Anchor" seems liable to confuse folks easily. Why not
TPTA to be clear? The two TA definitions following seem
to already allow for such confusion.

- 1.1: certificate-less TLS: what about DH-ANON? There's
no uniquely shared credential there. I thought you meant
TLS-PSK btw, but its turns out (in 3.3.3) that you really
mean TLS-SRP. Better to be clear about that up front.
(I'm not sure if TLS-SRP is really wise, but I guess this
is a corner case, right?)

- section 2, 2nd last para: this isn't really a "profile"
suggest s/profile/specification/

- 2.2.3: I think you could re-iterate that that better
take place via TLS here, or very bad things can happen:-)
(I know you say that in 3.2.3 but still...)

- 2.6: MAC addresses in CSR's seems a bit odd, maybe add
to the end of that sentence "...of an interface that the
EST CA is being asked to include in a public key
certificate" (but when does a cert include a MAC
address?) Seems like an odd example.

- 3.1, 2nd last para, last sentence: saying you MUST be
able to disable implicit TA databases seems likely to
mean that e.g. you couldn't do a client as a JS
application running in a browser or a browser plug-in. Is
that right and is that considered ok? Perhaps SHOULD
would be better there in terms of getting more/quicker
deployment? (3.2.3 also expresses this requirement.)

- 3.5 1st sentence: I think s/support/use/ here is right.
This is OPTIONAL to support but some servers might
REQUIRE it, which is bad for interop, but understandable
I guess. (Some server/CA policies will be such that not
all clients can work with 'em.)

- 3.5: Saying clients are RECOMMENDED to do this but that
PoP is OPTIONAL is confusing.

- 3.5: If tls-unique is MTI then you should say that in
section 3.3 somewhere. (The MUST verify for that in 3.5
makes it MTI for servers at least I think.)

- 3.5: Saying that you might need to hash the
challenge-password but not saying how seems wrong.  I'd
say just re-word to say that only <255 byte
challenge-passwords are supported here.

- 3.6.1: it seems odd to say that id-kp-cmcRA is used to
"authorize" an EST server.

- 4.1.1: (shameless self-promtion:-) If its useful you
could maybe make use of RFC 6920 for that out of band
confirmation. No problems if you'd rather not but the nih
format might be usable for that and might even help.
I expect you'll say that you've already implemented
something and its different, and that's fine.

- 4.1.1: what is a Publish Trust Anchor "control"?  I
don't recall the term control before.

- 4.2.1: I'm not clear from this what forms of PoP MUST
be implemented. Maybe 4.3 will make it clear.  (It
didn't.) I'd say maybe get rid of discussion of the more
complex forms of PoP from here.

- 4.2.2: is "rekeys the client certificate" sufficiently
clear?

- 4.2.2 and elsewhere: If human-readable content is
included, how do you know what language to use? Maybe
just say the EST server has to know somehow.

- 7, 2nd last para: typo: s/it is RECOMMEND/it is
RECOMMENDED/

- The secdir review [1] had a few comments that would be
good to address, (and I've not seen a response), in
particular, a generalisation of this seems worth a
mention in section 7: "In 2.6, the additional CSR
attributes are non-verifiable data.  E.g., if a client
puts a MAC address into its CSR, there is no way for the
EST server or CA to validate that input.  The client can
easily insert any data it wants into that request, which
can lead to MAC stealing or other impersonation attacks.
The CA should never put any normative information into
the certificate that it cannot validate from a trusted
source."

  [1] http://www.ietf.org/mail-archive/web/secdir/current/msg04029.html

[Ballot discuss]
D2. The way Section 3.2. is structured has ambiguities that seem likely to lead to interoperability problems.  It took me a couple of passes through Section 3.2 to realize that there are actually 5 different protocols here that have been glued together because they some superficial similarity.  It seems like it would be much clearer to organize this section in terms of the EST operations in Figure 5; for each operation, define the URI path it uses and the content types.  As it is, it's not immediately clear that you couldn't send a Full CMC message to an "enrollment of new clients" URI.

[Ballot comment]
This document is just mad for 2119 language. There are some places where I think they are used excessively, but I am trying to be on board with Stephen's philosophy of "if it's not going to mess up an implementation, ignore incorrect uses". There are no technical worries that I see. So, I will only mention two silly, more editorial points:

Section 2:

  This section does not include RFC 2119 key words.

Really? :-) I suggest deleting the sentence.

Section 3.1:

  The EST client is RECOMMENDED to be configured with...

I'm worried what the answer might be, but why didn't you say "The EST client SHOULD be configured with..." instead of contorting the adjective into a place perfectly suited for an auxiliary verb? (You've got similar things throughout the document: "It is RECOMMENDED that clients..." instead of "Clients SHOULD...".) Is there anything other than just a weird writing style that is making you use these strange passive sentences? It's not that you think SHOULD and RECOMMENDED mean two different things, right?

[Ballot comment]
Not sure I've seen any answer to Nevil's OPS DIR review:


I have reviewed this draft.  It "describes a simple yet functional
certificate management protocol targeting Public Key Infrastructure
(PKI) clients that need to acquire client certificates and associated
Certification Authority (CA) certificate(s)."

I'm certainly not a security expert, and this is a fine-detailed
document that's clearly intended for people who have a good understanding of public key infrastructure, in particular RFC 5274
(Certificate Management Messages over CMS).  Nonetheless, it's
well-written, and I had no difficulty understanding how an EST server
is intended to work.

When it comes to deploying an EST server, an organisation will obviously
need to understand how pkix works in considerable detail, and to make
sure that its various components will have to interwork properly with
each other.  That said, this draft document documents a new service -
as far as I can see - so anyone deploying it will need to be sure that
their existing pki infrastructure will support it properly.

I note that in several places (e.g. section 3.32.3) there's a reference
to "TLS 1.1 [RFC4346] or later versions."  That seems like a good idea,
allowing for newer versions of TLS to work with an EST server,  however
it does mean that EST implementors will need to check that their servers
continue to work properly with such newer versions.

Question: in section 1.1, Terminology, 'Implicit Trust Anchor:' gives
an example that includes "used by server's to authenticate
manufacturing installed client credentials."  Can you make it a bit
clearer just what that really means?


A few small typos:

section 3.4: s/the possesion and/the possesion of and/  ?

section 3.4  s/cipher suite pose such/cipher suite poses such/

section 3.7  /simpleenroll is split over a line break, leaving
            a lone / at the end of a line

section 4.2  s/verify the authorization the/
              verify the authorization of the/

section 4.2.1  s/used to for/used for/

sections 4.2.3 and 4.3.2  s/informative information/information/ ?

section 4.5.2  s/Csrattrs/csrattrs/  (why the uc C here?)

Cheers, Nevil

[Ballot discuss]
D1. The document requires the use of HTTP >=1.1.  Why is that the case?  It seems like HTTP/1.0 over TLS would be sufficient for what this document is trying to do.  Also, the queries in Section 3.2.2 would be valid HTTP/1.0, but are not with HTTP/1.1 (because they lack a Host field)

D2. The way Section 3.2. is structured has ambiguities that seem likely to lead to interoperability problems.  It took me a couple of passes through Section 3.2 to realize that there are actually 5 different protocols here that have been glued together because they some superficial similarity.  It seems like it would be much clearer to organize this section in terms of the EST operations in Figure 5; for each operation, define the URI path it uses and the content types.  As it is, it's not immediately clear that you couldn't send a Full CMC message to an "enrollment of new clients" URI.

[Ballot comment]
C1. The comment in Section 1 that EST doesn't define EST over UDP/DTLS seems a little non-sensical.

C2. In Section 3.2.1: This document doesn't "profile" the Content-Type header, it requires that specific values be used in it.  It seems like this section could be deleted in favor of Section 3.2.4.

C3. This may be obvious, but I don't see anywhere in the document that says that the EST application layer MUST NOT be used over bare HTTP, without TLS.  This might be good to note explicitly.

[Ballot comment]
This seems both explicit and wierdly non-specific.

  TLS cipher suites that include "_EXPORT_" and "_DES_" in their names
  MUST NOT be used.  These ciphers do not offer a sufficient level of
  protection; 40-bit crypto in 2013 doesn't offer acceptable protection
  and the use of DES is deprecated.

[Ballot discuss]
UPDATED: I've now sent the email for this twice, but Sean's not seeing it. Resending manually ...

My apologies - balloted this, but Sean hasn't seen the email yet. Just in case I managed to not send it at all, I'm sending it again.

The Discuss is not for the editors, but for some ADs ... I'll clear immediately after someone clubs me with a clue-by-four.

Dear APP and SEC ADs, in 1.  Introduction, there's this text:

  EST specifies how to transfer messages securely via HTTP over TLS
  (HTTPS) [RFC2818], where the HTTP headers and media types are used in
  conjunction with TLS.  HTTPS operates over TCP; this document does
  not specify EST over Datagram Transport Layer Security/User Datagram
  Protocol (DTLS/UDP). 

I wonder if the disclaimer about not specifying EST over DTLS/UDP is the right thing to say (it's not wrong, although I'm not sure why the parallel to EST/HTTP/TLS/TCP is EST/DTLS/UDP - isn't there a layer missing?).

Given that we already have CORE, which can be fairly characterized as an HTTP variant running over UDP, and given that we approved an IETF 87 BOF on "DTLS In Constrained Environments (DICE)" yesterday, EST over CORE/DTLS/UDP would be an reasonable proposal to think about in the not-too-distant future (not in the context of this document, of course!)

I know we're not going to speculate about possible future work that hasn't even been chartered yet in an RFC that lasts Internet-Forever, but if EST over DTLS/UDP is going to be mentioned at all, is there anything else that would be helpful to say?

At a minumum, I found myself wondering if "EST over DTLS/UDP" was excluded because of some deep security challenge that would prevent EST from ever working over CORE/DTLS/UDP, or just "silly rabbit, EST uses HTTP, and HTTP doesn't run over DTLS".

Even if there are obvious dragons in that direction, that may not be something to mention in the document, but if it is, I should probably ask now ...

[Ballot discuss]
My apologies - balloted this, but Sean hasn't seen the email yet. Just in case I managed to not send it at all, I'm sending it again.

The Discuss is not for the editors, but for some ADs ... I'll clear immediately after someone clubs me with a clue-by-four.

Dear APP and SEC ADs, in 1.  Introduction, there's this text:

  EST specifies how to transfer messages securely via HTTP over TLS
  (HTTPS) [RFC2818], where the HTTP headers and media types are used in
  conjunction with TLS.  HTTPS operates over TCP; this document does
  not specify EST over Datagram Transport Layer Security/User Datagram
  Protocol (DTLS/UDP). 

I wonder if the disclaimer about not specifying EST over DTLS/UDP is the right thing to say (it's not wrong, although I'm not sure why the parallel to EST/HTTP/TLS/TCP is EST/DTLS/UDP - isn't there a layer missing?).

Given that we already have CORE, which can be fairly characterized as an HTTP variant running over UDP, and given that we approved an IETF 87 BOF on "DTLS In Constrained Environments (DICE)" yesterday, EST over CORE/DTLS/UDP would be an reasonable proposal to think about in the not-too-distant future (not in the context of this document, of course!)

I know we're not going to speculate about possible future work that hasn't even been chartered yet in an RFC that lasts Internet-Forever, but if EST over DTLS/UDP is going to be mentioned at all, is there anything else that would be helpful to say?

At a minumum, I found myself wondering if "EST over DTLS/UDP" was excluded because of some deep security challenge that would prevent EST from ever working over CORE/DTLS/UDP, or just "silly rabbit, EST uses HTTP, and HTTP doesn't run over DTLS".

Even if there are obvious dragons in that direction, that may not be something to mention in the document, but if it is, I should probably ask now ...

[Ballot discuss]

I'll end up as a yes for this but I have a few things I'd
like to DISCUSS first:

(1) I don't get how the client discovers the
"arbitraryLabel1" etc in 3.2.1? Don't you need to say
that they're assumed to be known to the client? (Or
provide a discovery mechanism, which sounds like too much
work.) I'm also unsure of the point of using .well-known
with those labels, but can live with it.

(2) 3.2.1, .well-known URLs will often result in HTTP
re-direction. Don't you need to say more for HTTP
re-directions, e.g. that whatever is required for the
original server cert is also required to be true of the
location to which the HTTP client is re-directed? How
does the MUST for tls-unique values from the 1st TLS
session in 3.5 work with re-direction? I suspect that you
need to either work out how 3xx responses are done or
else explain it a bit more to me, because I didn't get
that it works now.

(3) 3.6.1: what does check "against" mean? Don't you need
to say more, e.g. URI has to be in a SAN of a TA or
something? (Similar question for 3.6.2.)

(4) 4.4.2 - is this the 1st mention of SMIME
Capabilities?  Don't you need it earlier if you use it
here? Maybe that's a cut'n'paste error?

[Ballot comment]

- intro, why call out TLS1.1 and have no reference to
5246? I think at least adding 5246 would be good since we
would like folks to use TLS1.2 for this (and for
everything). I'm fine if this says that it doesn't really
matter (in functional terms) what version of TLS you use
which is what I think you're trying to say.

- intro, would it be worthwhile also saying that this
doesn't have any relationship with 6712 (which updates
4210) and sort-of attempts to do the CMP-like equivalent
of this? The only reason to add soemthing like that would
be to offset reader confusion if they go look at 4210,
see its updated by 6712 and then are puzzled that this is
very different from 6712. (And yes, all involved, incl.
me, are sorry about the 20-th century screw up that lead
to us having both CMP and CMC;-)

- 1.1: TA as the abbreviation for "Third Party Trust
Anchor" seems liable to confuse folks easily. Why not
TPTA to be clear? The two TA definitions following seem
to already allow for such confusion.

- 1.1: certificate-less TLS: what about DH-ANON? There's
no uniquely shared credential there. I thought you meant
TLS-PSK btw, but its turns out (in 3.3.3) that you really
mean TLS-SRP. Better to be clear about that up front.
(I'm not sure if TLS-SRP is really wise, but I guess this
is a corner case, right?)

- section 2, 2nd last para: this isn't really a "profile"
suggest s/profile/specification/

- 2.2.3: I think you could re-iterate that that better
take place via TLS here, or very bad things can happen:-)
(I know you say that in 3.2.3 but still...)

- 2.6: MAC addresses in CSR's seems a bit odd, maybe add
to the end of that sentence "...of an interface that the
EST CA is being asked to include in a public key
certificate" (but when does a cert include a MAC
address?) Seems like an odd example.

- 3.1, 2nd last para, last sentence: saying you MUST be
able to disable implicit TA databases seems likely to
mean that e.g. you couldn't do a client as a JS
application running in a browser or a browser plug-in. Is
that right and is that considered ok? Perhaps SHOULD
would be better there in terms of getting more/quicker
deployment? (3.2.3 also expresses this requirement.)

- 3.5 1st sentence: I think s/support/use/ here is right.
This is OPTIONAL to support but some servers might
REQUIRE it, which is bad for interop, but understandable
I guess. (Some server/CA policies will be such that not
all clients can work with 'em.)

- 3.5: Saying clients are RECOMMENDED to do this but that
PoP is OPTIONAL is confusing.

- 3.5: If tls-unique is MTI then you should say that in
section 3.3 somewhere. (The MUST verify for that in 3.5
makes it MTI for servers at least I think.)

- 3.5: Saying that you might need to hash the
challenge-password but not saying how seems wrong.  I'd
say just re-word to say that only <255 byte
challenge-passwords are supported here.

- 3.6.1: it seems odd to say that id-kp-cmcRA is used to
"authorize" an EST server.

- 4.1.1: (shameless self-promtion:-) If its useful you
could maybe make use of RFC 6920 for that out of band
confirmation. No problems if you'd rather not but the nih
format might be usable for that and might even help.
I expect you'll say that you've already implemented
something and its different, and that's fine.

- 4.1.1: what is a Publish Trust Anchor "control"?  I
don't recall the term control before.

- 4.2.1: I'm not clear from this what forms of PoP MUST
be implemented. Maybe 4.3 will make it clear.  (It
didn't.) I'd say maybe get rid of discussion of the more
complex forms of PoP from here.

- 4.2.2: is "rekeys the client certificate" sufficiently
clear?

- 4.2.2 and elsewhere: If human-readable content is
included, how do you know what language to use? Maybe
just say the EST server has to know somehow.

- 7, 2nd last para: typo: s/it is RECOMMEND/it is
RECOMMENDED/

- The secdir review [1] had a few comments that would be
good to address, (and I've not seen a response), in
particular, a generalisation of this seems worth a
mention in section 7: "In 2.6, the additional CSR
attributes are non-verifiable data.  E.g., if a client
puts a MAC address into its CSR, there is no way for the
EST server or CA to validate that input.  The client can
easily insert any data it wants into that request, which
can lead to MAC stealing or other impersonation attacks.
The CA should never put any normative information into
the certificate that it cannot validate from a trusted
source."

  [1] http://www.ietf.org/mail-archive/web/secdir/current/msg04029.html

Reddy, et al.            Expires August 2, 2019                [Page 3]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

1.  Introduction

  A distributed denial-of-service (DDoS) attack is a distributed
  attempt to make machines or network resources unavailable to their
  intended users.  In most cases, sufficient scale for an effective
  attack can be achieved by compromising enough end-hosts and using
  those infected hosts to perpetrate and amplify the attack.  The
  victim in this attack can be an application server, a host, a router,
  a firewall, or an entire network.

  Network applications have finite resources like CPU cycles, the
  number of processes or threads they can create and use, the maximum
  number of simultaneous connections they can handle, the limited
  resources of the control plane, etc.  When processing network
  traffic, such applications are supposed to use these resources to
  provide the intended functionality in the most efficient manner.
  However, a DDoS attacker may be able to prevent an application from
  performing its intended task by making the application exhaust its
  finite resources.

  A TCP DDoS SYN-flood [RFC4987], for example, is a memory-exhausting
  attack while an ACK-flood is a CPU-exhausting attack.  Attacks on the
  link are carried out by sending enough traffic so that the link
  becomes congested, thereby likely causing packet loss for legitimate
  traffic.  Stateful firewalls can also be attacked by sending traffic
  that causes the firewall to maintain an excessive number of states
  that may jeopardize the firewall's operation overall, besides likely
  performance impacts.  The firewall then runs out of memory, and can
  no longer instantiate the states required to process legitimate
  flows.  Other possible DDoS attacks are discussed in [RFC4732].

  In many cases, it may not be possible for network administrators to
  determine the cause(s) of an attack.  They may instead just realize
  that certain resources seem to be under attack.  This document
  defines a lightweight protocol that allows a DOTS client to request
  mitigation from one or more DOTS servers for protection against
  detected, suspected, or anticipated attacks.  This protocol enables
  cooperation between DOTS agents to permit a highly-automated network
  defense that is robust, reliable, and secure.  Note that "secure"
  means the support of the features defined in Section 2.4 of
  [I-D.ietf-dots-requirements].

  An example of a network diagram that illustrates a deployment of DOTS
  agents is shown in Figure 1.  In this example, a DOTS server is
  operating on the access network.  A DOTS client is located on the LAN
  (Local Area Network), while a DOTS gateway is embedded in the CPE
  (Customer Premises Equipment).

Reddy, et al.            Expires August 2, 2019                [Page 4]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      Network
      Resource        CPE router        Access network    __________
    +-----------+  +--------------+    +-------------+    /          \
    |          |___|              |____|            |___ | Internet |
    |DOTS client|  | DOTS gateway |    | DOTS server |    |          |
    |          |  |              |    |            |    |          |
    +-----------+  +--------------+    +-------------+    \__________/

                  Figure 1: Sample DOTS Deployment (1)

  DOTS servers can also be reachable over the Internet, as depicted in
  Figure 2.

      Network                                          DDoS mitigation
      Resource          CPE router      __________        service
    +-----------+  +-------------+    /          \    +-------------+
    |          |___|            |____|          |___ |            |
    |DOTS client|  |DOTS gateway |    | Internet |    | DOTS server |
    |          |  |            |    |          |    |            |
    +-----------+  +-------------+    \__________/    +-------------+

                  Figure 2: Sample DOTS Deployment (2)

  In typical deployments, the DOTS client belongs to a different
  administrative domain than the DOTS server.  For example, the DOTS
  client is embedded in a firewall protecting services owned and
  operated by a customer, while the DOTS server is owned and operated
  by a different administrative entity (service provider, typically)
  providing DDoS mitigation services.  The latter might or might not
  provide connectivity services to the network hosting the DOTS client.

  The DOTS server may (not) be co-located with the DOTS mitigator.  In
  typical deployments, the DOTS server belongs to the same
  administrative domain as the mitigator.  The DOTS client can
  communicate directly with a DOTS server or indirectly via a DOTS
  gateway.

  The document adheres to the DOTS architecture
  [I-D.ietf-dots-architecture].  The requirements for DOTS signal
  channel protocol are documented in [I-D.ietf-dots-requirements].
  This document satisfies all the use cases discussed in
  [I-D.ietf-dots-use-cases].

  This document focuses on the DOTS signal channel.  This is a
  companion document of the DOTS data channel specification
  [I-D.ietf-dots-data-channel] that defines a configuration and a bulk
  data exchange mechanism supporting the DOTS signal channel.

Reddy, et al.            Expires August 2, 2019                [Page 5]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

2.  Terminology

  The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
  "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
  "OPTIONAL" in this document are to be interpreted as described in BCP
  14 [RFC2119][RFC8174] when, and only when, they appear in all
  capitals, as shown here.

  (D)TLS is used for statements that apply to both Transport Layer
  Security [RFC5246][RFC8446] and Datagram Transport Layer Security
  [RFC6347].  Specific terms are used for any statement that applies to
  either protocol alone.

  The reader should be familiar with the terms defined in
  [I-D.ietf-dots-requirements].

  The meaning of the symbols in YANG tree diagrams is defined in
  [RFC8340].

3.  Design Overview

  The DOTS signal channel is built on top of the Constrained
  Application Protocol (CoAP) [RFC7252], a lightweight protocol
  originally designed for constrained devices and networks.  The many
  features of CoAP (expectation of packet loss, support for
  asynchronous Non-confirmable messaging, congestion control, small
  message overhead limiting the need for fragmentation, use of minimal
  resources, and support for (D)TLS) makes it a good candidate to build
  the DOTS signaling mechanism from.

  The DOTS signal channel is layered on existing standards (Figure 3).

                          +---------------------+
                          | DOTS Signal Channel |
                          +---------------------+
                          |        CoAP        |
                          +----------+----------+
                          |  TLS    |  DTLS  |
                          +----------+----------+
                          |  TCP    |  UDP    |
                          +----------+----------+
                          |          IP        |
                          +---------------------+

    Figure 3: Abstract Layering of DOTS Signal Channel over CoAP over
                                  (D)TLS

Reddy, et al.            Expires August 2, 2019                [Page 6]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  By default, a DOTS signal channel MUST run over port number TBD as
  defined in Section 9.1, for both UDP and TCP, unless the DOTS server
  has a mutual agreement with its DOTS clients to use a different port
  number.  DOTS clients MAY alternatively support means to dynamically
  discover the ports used by their DOTS servers (e.g.,
  [I-D.boucadair-dots-server-discovery]).  In order to use a distinct
  port number (as opposed to TBD), DOTS clients and servers SHOULD
  support a configurable parameter to supply the port number to use.
  The rationale for not using the default port number 5684 ((D)TLS
  CoAP) is to allow for differentiated behaviors in environments where
  both a DOTS gateway and an IoT gateway (e.g., Figure 3 of [RFC7452])
  are present.

  The signal channel uses the "coaps" URI scheme defined in Section 6
  of [RFC7252] and the "coaps+tcp" URI scheme defined in Section 8.2 of
  [RFC8323] to identify DOTS server resources accessible using CoAP
  over UDP secured with DTLS and CoAP over TCP secured with TLS,
  respectively.

  The signal channel is initiated by the DOTS client (Section 4.4).
  Once the signal channel is established, the DOTS agents periodically
  send heartbeats to keep the channel active (Section 4.7).  At any
  time, the DOTS client may send a mitigation request message to a DOTS
  server over the active channel.  While mitigation is active (because
  of the higher likelihood of packet loss during a DDoS attack), the
  DOTS server periodically sends status messages to the client,
  including basic mitigation feedback details.  Mitigation remains
  active until the DOTS client explicitly terminates mitigation, or the
  mitigation lifetime expires.

  DOTS signaling can happen with DTLS over UDP and TLS over TCP.
  Likewise, DOTS requests may be sent using IPv4 or IPv6 transfer
  capabilities.  A Happy Eyeballs procedure for DOTS signal channel is
  specified in Section 4.3.

  Messages exchanged between DOTS agents are serialized using Concise
  Binary Object Representation (CBOR) [RFC7049], a binary encoding
  scheme designed for small code and message size.  CBOR-encoded
  payloads are used to carry signal channel-specific payload messages
  which convey request parameters and response information such as
  errors.  In order to allow reusing data models across protocols,
  [RFC7951] specifies the JavaScript Object Notation (JSON) encoding of
  YANG-modeled data.  A similar effort for CBOR is defined in
  [I-D.ietf-core-yang-cbor].

  DOTS agents primarily determine that a CBOR data structure is a DOTS
  signal channel object from the application context, such as from the
  port number assigned to the DOTS signal channel.  The other method

Reddy, et al.            Expires August 2, 2019                [Page 7]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  DOTS agents use to indicate that a CBOR data structure is a DOTS
  signal channel object is the use of the "application/dots+cbor"
  content type (Section 9.3).

  This document specifies a YANG module for representing DOTS
  mitigation scopes, DOTS signal channel session configuration data,
  and DOTS redirected signalling (Section 5).  Representing these data
  as CBOR data can either follow the rules in [I-D.ietf-core-yang-cbor]
  or those in [RFC7951] combined with JSON/CBOR conversion rules in
  [RFC7049]; both approaches produce a valid encoding.  All parameters
  in the payload of the DOTS signal channel are mapped to CBOR types as
  specified in Section 6.

  In order to prevent fragmentation, DOTS agents must follow the
  recommendations documented in Section 4.6 of [RFC7252].  Refer to
  Section 7.3 for more details.

  DOTS agents MUST support GET, PUT, and DELETE CoAP methods.  The
  payload included in CoAP responses with 2.xx Response Codes MUST be
  of content type "application/dots+cbor".  CoAP responses with 4.xx
  and 5.xx error Response Codes MUST include a diagnostic payload
  (Section 5.5.2 of [RFC7252]).  The Diagnostic Payload may contain
  additional information to aid troubleshooting.

  In deployments where multiple DOTS clients are enabled in a network
  (owned and operated by the same entity), the DOTS server may detect
  conflicting mitigation requests from these clients.  This document
  does not aim to specify a comprehensive list of conditions under
  which a DOTS server will characterize two mitigation requests from
  distinct DOTS clients as conflicting, nor recommend a DOTS server
  behavior for processing conflicting mitigation requests.  Those
  considerations are implementation- and deployment-specific.
  Nevertheless, the document specifies the mechanisms to notify DOTS
  clients when conflicts occur, including the conflict cause
  (Section 4.4).

  In deployments where one or more translators (e.g., Traditional NAT
  [RFC3022], CGN [RFC6888], NAT64 [RFC6146], NPTv6 [RFC6296]) are
  enabled between the client's network and the DOTS server, DOTS signal
  channel messages forwarded to a DOTS server MUST NOT include internal
  IP addresses/prefixes and/or port numbers; external addresses/
  prefixes and/or port numbers as assigned by the translator MUST be
  used instead.  This document does not make any recommendation about
  possible translator discovery mechanisms.  The following are some
  (non-exhaustive) deployment examples that may be considered:

  o  Port Control Protocol (PCP) [RFC6887] or Session Traversal
      Utilities for NAT (STUN) [RFC5389] may be used to retrieve the

Reddy, et al.            Expires August 2, 2019                [Page 8]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      external addresses/prefixes and/or port numbers.  Information
      retrieved by means of PCP or STUN will be used to feed the DOTS
      signal channel messages that will be sent to a DOTS server.

  o  A DOTS gateway may be co-located with the translator.  The DOTS
      gateway will need to update the DOTS messages, based upon the
      local translator's binding table.

4.  DOTS Signal Channel: Messages & Behaviors

4.1.  DOTS Server(s) Discovery

  This document assumes that DOTS clients are provisioned with the
  reachability information of their DOTS server(s) using any of a
  variety of means (e.g., local configuration, or dynamic means such as
  DHCP).  The description of such means is out of scope of this
  document.

  Likewise, it is out of scope of this document to specify the behavior
  to be followed by a DOTS client to send DOTS requests when multiple
  DOTS servers are provisioned (e.g., contact all DOTS servers, select
  one DOTS server among the list).

4.2.  CoAP URIs

  The DOTS server MUST support the use of the path-prefix of "/.well-
  known/" as defined in [RFC5785] and the registered name of "dots".
  Each DOTS operation is indicated by a path-suffix that indicates the
  intended operation.  The operation path (Table 1) is appended to the
  path-prefix to form the URI used with a CoAP request to perform the
  desired DOTS operation.

        +-----------------------+----------------+-------------+
        | Operation            | Operation Path | Details    |
        +-----------------------+----------------+-------------+
        | Mitigation            | /mitigate      | Section 4.4 |
        +-----------------------+----------------+-------------+
        | Session configuration | /config        | Section 4.5 |
        +-----------------------+----------------+-------------+

            Table 1: Operations and their Corresponding URIs

4.3.  Happy Eyeballs for DOTS Signal Channel

  [I-D.ietf-dots-requirements] mentions that DOTS agents will have to
  support both connectionless and connection-oriented protocols.  As
  such, the DOTS signal channel is designed to operate with DTLS over
  UDP and TLS over TCP.  Further, a DOTS client may acquire a list of

Reddy, et al.            Expires August 2, 2019                [Page 9]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  IPv4 and IPv6 addresses (Section 4.1), each of which can be used to
  contact the DOTS server using UDP and TCP.  The following specifies
  the procedure to follow to select the address family and the
  transport protocol for sending DOTS signal channel messages.

  Such procedure is needed to avoid experiencing long connection
  delays.  For example, if an IPv4 path to reach a DOTS server is
  functional, but the DOTS server's IPv6 path is non-functional, a
  dual-stack DOTS client may experience a significant connection delay
  compared to an IPv4-only DOTS client, in the same network conditions.
  The other problem is that if a middlebox between the DOTS client and
  DOTS server is configured to block UDP traffic, the DOTS client will
  fail to establish a DTLS association with the DOTS server and, as a
  consequence, will have to fall back to TLS over TCP, thereby
  incurring significant connection delays.

  To overcome these connection setup problems, the DOTS client attempts
  to connect to its DOTS server(s) using both IPv6 and IPv4, and tries
  both DTLS over UDP and TLS over TCP in a manner similar to the Happy
  Eyeballs mechanism [RFC8305].  These connection attempts are
  performed by the DOTS client when it initializes, or in general when
  it has to select an address family and transport to contact its DOTS
  server.  The results of the Happy Eyeballs procedure are used by the
  DOTS client for sending its subsequent messages to the DOTS server.
  Note that the DOTS client after successfully establishing a
  connection MUST cache information regarding the outcome of each
  connection attempt for a specific time period, and it uses that
  information to avoid thrashing the network with subsequent attempts.

  The order of preference of the DOTS signal channel address family and
  transport protocol (most preferred first) is: UDP over IPv6, UDP over
  IPv4, TCP over IPv6, and finally TCP over IPv4.  This order adheres
  to the address preference order specified in [RFC6724] and the DOTS
  signal channel preference which privileges the use of UDP over TCP
  (to avoid TCP's head of line blocking).

  In reference to Figure 4, the DOTS client sends two TCP SYNs and two
  DTLS ClientHello messages at the same time over IPv6 and IPv4.  In
  this example, it is assumed that the IPv6 path is broken and UDP
  traffic is dropped by a middlebox but has little impact to the DOTS
  client because there is no long delay before using IPv4 and TCP.  The
  DOTS client periodically repeats the mechanism to discover whether
  DOTS signal channel messages with DTLS over UDP becomes available
  from the DOTS server, so the DOTS client can migrate the DOTS signal
  channel from TCP to UDP.  Such probing SHOULD NOT be done more
  frequently than every 24 hours and MUST NOT be done more frequently
  than every 5 minutes.

Reddy, et al.            Expires August 2, 2019                [Page 10]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  A single DOTS signal channel between DOTS agents can be used to
  exchange multiple DOTS signal messages.  To reduce DOTS client and
  DOTS server workload, DOTS clients SHOULD re-use the (D)TLS session.

  +-----------+                                          +-----------+
  |DOTS client|                                          |DOTS server|
  +-----------+                                          +-----------+
        |                                                      |
        |--DTLS ClientHello, IPv6 ---->X                        |
        |--TCP SYN, IPv6-------------->X                        |
        |--DTLS ClientHello, IPv4 ---->X                        |
        |--TCP SYN, IPv4--------------------------------------->|
        |--DTLS ClientHello, IPv6 ---->X                        |
        |--TCP SYN, IPv6-------------->X                        |
        |<-TCP SYNACK-------------------------------------------|
        |--DTLS ClientHello, IPv4 ---->X                        |
        |--TCP ACK--------------------------------------------->|
        |<------------Establish TLS Session-------------------->|
        |----------------DOTS signal--------------------------->|
        |                                                      |

                Figure 4: DOTS Happy Eyeballs (Sample Flow)

4.4.  DOTS Mitigation Methods

  The following methods are used by a DOTS client to request, withdraw,
  or retrieve the status of mitigation requests:

  PUT:    DOTS clients use the PUT method to request mitigation from a
          DOTS server (Section 4.4.1).  During active mitigation, DOTS
          clients may use PUT requests to carry mitigation efficacy
          updates to the DOTS server (Section 4.4.3).

  GET:    DOTS clients may use the GET method to subscribe to DOTS
          server status messages, or to retrieve the list of its
          mitigations maintained by a DOTS server (Section 4.4.2).

  DELETE: DOTS clients use the DELETE method to withdraw a request for
          mitigation from a DOTS server (Section 4.4.4).

  Mitigation request and response messages are marked as Non-
  confirmable messages (Section 2.2 of [RFC7252]).

  DOTS agents SHOULD follow the data transmission guidelines discussed
  in Section 3.1.3 of [RFC8085] and control transmission behavior by
  not sending more than one UDP datagram per round-trip time (RTT) to
  the peer DOTS agent on average.

Reddy, et al.            Expires August 2, 2019                [Page 11]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  Requests marked by the DOTS client as Non-confirmable messages are
  sent at regular intervals until a response is received from the DOTS
  server.  If the DOTS client cannot maintain an RTT estimate, it
  SHOULD NOT send more than one Non-confirmable request every 3
  seconds, and SHOULD use an even less aggressive rate whenever
  possible (case 2 in Section 3.1.3 of [RFC8085]).

  JSON diagnostic notation is used to illustrate the various methods
  defined in the following sub-sections.  Also, the examples use the
  Labels defined in Sections 9.6.2, 9.6.3, 9.6.4, and 9.6.5.

4.4.1.  Request Mitigation

  When a DOTS client requires mitigation for some reason, the DOTS
  client uses the CoAP PUT method to send a mitigation request to its
  DOTS server(s) (Figures 5 and 6).

  If a DOTS client is entitled to solicit the DOTS service, the DOTS
  server enables mitigation on behalf of the DOTS client by
  communicating the DOTS client's request to a mitigator (which may be
  colocated with the DOTS server) and relaying the feedback of the
  thus-selected mitigator to the requesting DOTS client.

    Header: PUT (Code=0.03)
    Uri-Path: ".well-known"
    Uri-Path: "dots"
    Uri-Path: "mitigate"
    Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
    Uri-Path: "mid=123"
    Content-Format: "application/dots+cbor"
    {
      ...
    }

            Figure 5: PUT to Convey DOTS Mitigation Requests

  The order of the Uri-Path options is important as it defines the CoAP
  resource.  In particular, 'mid' MUST follow 'cuid'.

  The additional Uri-Path parameters to those defined in Section 4.2
  are as follows:

  cuid:  Stands for Client Unique Identifier.  A globally unique
      identifier that is meant to prevent collisions among DOTS clients,
      especially those from the same domain.  It MUST be generated by
      DOTS clients.

Reddy, et al.            Expires August 2, 2019                [Page 12]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      Implementations SHOULD set 'cuid' to the output of a cryptographic
      hash algorithm whose input is the Distinguished Encoding Rules
      (DER)-encoded Abstract Syntax Notation One (ASN.1) representation
      of the Subject Public Key Info (SPKI) of the DOTS client X.509
      certificate [RFC5280], the DOTS client raw public key [RFC7250],
      or the "Pre-Shared Key (PSK) identity" used by the DOTS client in
      the TLS ClientKeyExchange message.  In this version of the
      specification, the cryptographic hash algorithm used is SHA-256
      [RFC6234].  The output of the cryptographic hash algorithm is
      truncated to 16 bytes; truncation is done by stripping off the
      final 16 bytes.  The truncated output is base64url encoded.

      The 'cuid' is intended to be stable when communicating with a
      given DOTS server, i.e., the 'cuid' used by a DOTS client SHOULD
      NOT change over time.  Distinct 'cuid' values MAY be used by a
      single DOTS client per DOTS server.

      DOTS servers MUST return 4.09 (Conflict) error code to a DOTS peer
      to notify that the 'cuid' is already in-use by another DOTS
      client.  Upon receipt of that error code, a new 'cuid' MUST be
      generated by the DOTS peer (e.g., using [RFC4122]).

      Client-domain DOTS gateways MUST handle 'cuid' collision directly
      and it is RECOMMENDED that 'cuid' collision is handled directly by
      server-domain DOTS gateways.

      DOTS gateways MAY rewrite the 'cuid' used by peer DOTS clients.
      Triggers for such rewriting are out of scope.

      This is a mandatory Uri-Path parameter.

  mid:  Identifier for the mitigation request represented with an
      integer.  This identifier MUST be unique for each mitigation
      request bound to the DOTS client, i.e., the 'mid' parameter value
      in the mitigation request needs to be unique (per 'cuid' and DOTS
      server) relative to the 'mid' parameter values of active
      mitigation requests conveyed from the DOTS client to the DOTS
      server.

      In order to handle out-of-order delivery of mitigation requests,
      'mid' values MUST increase monotonically.

      If the 'mid' value has reached 3/4 of (2**32 - 1) (i.e.,
      3221225471) and no attack is detected, the DOTS client MUST reset
      'mid' to 0 to handle 'mid' rollover.  If the DOTS client maintains
      mitigation requests with pre-configured scopes, it MUST re-create
      them with the 'mid' restarting at 0.

Reddy, et al.            Expires August 2, 2019                [Page 13]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      This identifier MUST be generated by the DOTS client.

      This is a mandatory Uri-Path parameter.

  'cuid' and 'mid' MUST NOT appear in the PUT request message body
  (Figure 6).

    Content-Format: "application/dots+cbor"
    {
      "ietf-dots-signal-channel:mitigation-scope": {
        "scope": [
          {
            "target-prefix": [
                "string"
              ],
            "target-port-range": [
                {
                  "lower-port": number,
                  "upper-port": number
                }
              ],
              "target-protocol": [
                number
              ],
              "target-fqdn": [
                "string"
              ],
              "target-uri": [
                "string"
              ],
              "alias-name": [
                "string"
              ],
            "lifetime": number,
            "trigger-mitigation": true|false
          }
        ]
      }
    }

      Figure 6: PUT to Convey DOTS Mitigation Requests (Message Body
                                  Schema)

  The parameters in the CBOR body (Figure 6) of the PUT request are
  described below:

Reddy, et al.            Expires August 2, 2019                [Page 14]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  target-prefix:  A list of prefixes identifying resources under
      attack.  Prefixes are represented using Classless Inter-Domain
      Routing (CIDR) notation [RFC4632].
      As a reminder, the prefix length must be less than or equal to 32
      (resp. 128) for IPv4 (resp.  IPv6).

      The prefix list MUST NOT include broadcast, loopback, or multicast
      addresses.  These addresses are considered as invalid values.  In
      addition, the DOTS server MUST validate that target prefixes are
      within the scope of the DOTS client's domain.  Other validation
      checks may be supported by DOTS servers.

      This is an optional attribute.

  target-port-range:  A list of port numbers bound to resources under
      attack.

      A port range is defined by two bounds, a lower port number (lower-
      port) and an upper port number (upper-port).  When only 'lower-
      port' is present, it represents a single port number.

      For TCP, UDP, Stream Control Transmission Protocol (SCTP)
      [RFC4960], or Datagram Congestion Control Protocol (DCCP)
      [RFC4340], a range of ports can be, for example, 0-1023,
      1024-65535, or 1024-49151.

      This is an optional attribute.

  target-protocol:  A list of protocols involved in an attack.  Values
      are taken from the IANA protocol registry [proto_numbers].

      If 'target-protocol' is not specified, then the request applies to
      any protocol.

      This is an optional attribute.

  target-fqdn:  A list of Fully Qualified Domain Names (FQDNs)
      identifying resources under attack [RFC8499].

      How a name is passed to an underlying name resolution library is
      implementation- and deployment-specific.  Nevertheless, once the
      name is resolved into one or multiple IP addresses, DOTS servers
      MUST apply the same validation checks as those for 'target-
      prefix'.

      The use of FQDNs may be suboptimal because:

Reddy, et al.            Expires August 2, 2019                [Page 15]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      *  It induces both an extra load and increased delays on the DOTS
        server to handle and manage DNS resolution requests.

      *  It does not guarantee that the DOTS server will resolve a name
        to the same IP addresses that the DOTS client does.

      This is an optional attribute.

  target-uri:  A list of Uniform Resource Identifiers (URIs) [RFC3986]
      identifying resources under attack.

      The same validation checks used for 'target-fqdn' MUST be followed
      by DOTS servers to validate a target URI.

      This is an optional attribute.

  alias-name:  A list of aliases of resources for which the mitigation
      is requested.  Aliases can be created using the DOTS data channel
      (Section 6.1 of [I-D.ietf-dots-data-channel]), direct
      configuration, or other means.

      An alias is used in subsequent signal channel exchanges to refer
      more efficiently to the resources under attack.

      This is an optional attribute.

  lifetime:  Lifetime of the mitigation request in seconds.  The
      RECOMMENDED lifetime of a mitigation request is 3600 seconds --
      this value was chosen to be long enough so that refreshing is not
      typically a burden on the DOTS client, while still making the
      request expire in a timely manner when the client has unexpectedly
      quit.  DOTS clients MUST include this parameter in their
      mitigation requests.  Upon the expiry of this lifetime, and if the
      request is not refreshed, the mitigation request is removed.  The
      request can be refreshed by sending the same request again.

      A lifetime of '0' in a mitigation request is an invalid value.

      A lifetime of negative one (-1) indicates indefinite lifetime for
      the mitigation request.  The DOTS server MAY refuse indefinite
      lifetime, for policy reasons; the granted lifetime value is
      returned in the response.  DOTS clients MUST be prepared to not be
      granted mitigations with indefinite lifetimes.

      The DOTS server MUST always indicate the actual lifetime in the
      response and the remaining lifetime in status messages sent to the
      DOTS client.

Reddy, et al.            Expires August 2, 2019                [Page 16]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      This is a mandatory attribute.

  trigger-mitigation:  If the parameter value is set to 'false', DDoS
      mitigation will not be triggered for the mitigation request unless
      the DOTS signal channel session is lost.

      If the DOTS client ceases to respond to heartbeat messages, the
      DOTS server can detect that the DOTS signal channel session is
      lost.

      The default value of the parameter is 'true' (that is, the
      mitigation starts immediately).  If 'trigger-mitigation' is not
      present in a request, this is equivalent to receiving a request
      with 'trigger-mitigation' set to 'true'.

      This is an optional attribute.

  In deployments where server-domain DOTS gateways are enabled,
  identity information about the origin source client domain SHOULD be
  propagated to the DOTS server.  That information is meant to assist
  the DOTS server to enforce some policies such as grouping DOTS
  clients that belong to the same DOTS domain, limiting the number of
  DOTS requests, and identifying the mitigation scope.  These policies
  can be enforced per-client, per-client domain, or both.  Also, the
  identity information may be used for auditing and debugging purposes.

  Figure 7 shows an example of a request relayed by a server-domain
  DOTS gateway.

    Header: PUT (Code=0.03)
    Uri-Path: ".well-known"
    Uri-Path: "dots"
    Uri-Path: "mitigate"
    Uri-Path: "cdid=7eeaf349529eb55ed50113"
    Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
    Uri-Path: "mid=123"
    Content-Format: "application/dots+cbor"
    {
      ...
    }

      Figure 7: PUT for DOTS Mitigation Request as Relayed by a DOTS
                                  Gateway

  A server-domain DOTS gateway SHOULD add the following Uri-Path
  parameter:

Reddy, et al.            Expires August 2, 2019                [Page 17]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  cdid:  Stands for Client Domain Identifier.  The 'cdid' is conveyed
      by a server-domain DOTS gateway to propagate the source domain
      identity from the gateway's client-facing-side to the gateway's
      server-facing-side, and from the gateway's server-facing-side to
      the DOTS server. 'cdid' may be used by the final DOTS server for
      policy enforcement purposes (e.g., enforce a quota on filtering
      rules).  These policies are deployment-specific.

      Server-domain DOTS gateways SHOULD support a configuration option
      to instruct whether 'cdid' parameter is to be inserted.

      In order to accommodate deployments that require enforcing per-
      client policies, per-client domain policies, or a combination
      thereof, server-domain DOTS gateways instructed to insert the
      'cdid' parameter MUST supply the SPKI hash of the DOTS client
      X.509 certificate, the DOTS client raw public key, or the hash of
      the "PSK identity" in the 'cdid', following the same rules for
      generating the hash conveyed in 'cuid', which is then used by the
      ultimate DOTS server to determine the corresponding client's
      domain.  The 'cdid' generated by a server-domain gateway is likely
      to be the same as the 'cuid' except if the DOTS message was
      relayed by a client-domain DOTS gateway or the 'cuid' was
      generated from a rogue DOTS client.

      If a DOTS client is provisioned, for example, with distinct
      certificates as a function of the peer server-domain DOTS gateway,
      distinct 'cdid' values may be supplied by a server-domain DOTS
      gateway.  The ultimate DOTS server MUST treat those 'cdid' values
      as equivalent.

      The 'cdid' attribute MUST NOT be generated and included by DOTS
      clients.

      DOTS servers MUST ignore 'cdid' attributes that are directly
      supplied by source DOTS clients or client-domain DOTS gateways.
      This implies that first server-domain DOTS gateways MUST strip
      'cdid' attributes supplied by DOTS clients.  DOTS servers SHOULD
      support a configuration parameter to identify DOTS gateways that
      are trusted to supply 'cdid' attributes.

      Only single-valued 'cdid' are defined in this document.

      This is an optional Uri-Path.  When present, 'cdid' MUST be
      positioned before 'cuid'.

  A DOTS gateway MAY add the CoAP Hop-Limit Option
  [I-D.ietf-core-hop-limit].

Reddy, et al.            Expires August 2, 2019                [Page 18]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  Because of the complexity to handle partial failure cases, this
  specification does not allow for including multiple mitigation
  requests in the same PUT request.  Concretely, a DOTS client MUST NOT
  include multiple entries in the 'scope' array of the same PUT
  request.

  FQDN and URI mitigation scopes may be thought of as a form of scope
  alias, in which the addresses associated with the domain name or URI
  (as resolved by the DOTS server) represent the full scope of the
  mitigation.

  In the PUT request at least one of the attributes 'target-prefix',
  'target-fqdn','target-uri', or 'alias-name' MUST be present.

  Attributes and Uri-Path parameters with empty values MUST NOT be
  present in a request and render the entire request invalid.

  Figure 8 shows a PUT request example to signal that TCP port numbers
  80, 8080, and 443 used by 2001:db8:6401::1 and 2001:db8:6401::2
  servers are under attack (illustrated in JSON diagnostic notation).
  The presence of 'cdid' indicates that a server-domain DOTS gateway
  has modified the initial PUT request sent by the DOTS client.  Note
  that 'cdid' MUST NOT appear in the PUT request message body.

Reddy, et al.            Expires August 2, 2019                [Page 19]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

    Header: PUT (Code=0.03)
    Uri-Path: ".well-known"
    Uri-Path: "dots"
    Uri-Path: "mitigate"
    Uri-Path: "cdid=7eeaf349529eb55ed50113"
    Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
    Uri-Path: "mid=123"
    Content-Format: "application/dots+cbor"
    {
      "ietf-dots-signal-channel:mitigation-scope": {
        "scope": [
          {
            "target-prefix": [
                "2001:db8:6401::1/128",
                "2001:db8:6401::2/128"
              ],
            "target-port-range": [
              {
                "lower-port": 80
              },
              {
                "lower-port": 443
              },
              {
                  "lower-port": 8080
              }
              ],
              "target-protocol": [
                6
              ],
            "lifetime": 3600
          }
        ]
      }
    }

          Figure 8: PUT for DOTS Mitigation Request (An Example)

  The corresponding CBOR encoding format for the payload is shown in
  Figure 9.

Reddy, et al.            Expires August 2, 2019                [Page 20]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  A1                                      # map(1)
      01                                  # unsigned(1)
      A1                                  # map(1)
        02                                # unsigned(2)
        81                                # array(1)
            A3                            # map(3)
              06                          # unsigned(6)
              82                          # array(2)
                  74                      # text(20)
                    323030313A6462383A363430313A3A312F313238
                  74                      # text(20)
                    323030313A6462383A363430313A3A322F313238
              07                          # unsigned(7)
              83                          # array(3)
                  A1                      # map(1)
                    08                    # unsigned(8)
                    18 50                # unsigned(80)
                  A1                      # map(1)
                    08                    # unsigned(8)
                    19 01BB              # unsigned(443)
                  A1                      # map(1)
                    08                    # unsigned(8)
                    19 1F90              # unsigned(8080)
              0A                          # unsigned(10)
              81                          # array(1)
                  06                      # unsigned(6)
              0E                          # unsigned(14)
                  19 0E10                  # unsigned(3600)

            Figure 9: PUT for DOTS Mitigation Request (CBOR)

  In both DOTS signal and data channel sessions, the DOTS client MUST
  authenticate itself to the DOTS server (Section 8).  The DOTS server
  MAY use the algorithm presented in Section 7 of [RFC7589] to derive
  the DOTS client identity or username from the client certificate.
  The DOTS client identity allows the DOTS server to accept mitigation
  requests with scopes that the DOTS client is authorized to manage.

  The DOTS server couples the DOTS signal and data channel sessions
  using the DOTS client identity and optionally the 'cdid' parameter
  value, so the DOTS server can validate whether the aliases conveyed
  in the mitigation request were indeed created by the same DOTS client
  using the DOTS data channel session.  If the aliases were not created
  by the DOTS client, the DOTS server MUST return 4.00 (Bad Request) in
  the response.

  The DOTS server couples the DOTS signal channel sessions using the
  DOTS client identity and optionally the 'cdid' parameter value, and

Reddy, et al.            Expires August 2, 2019                [Page 21]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  the DOTS server uses 'mid' and 'cuid' Uri-Path parameter values to
  detect duplicate mitigation requests.  If the mitigation request
  contains the 'alias-name' and other parameters identifying the target
  resources (such as 'target-prefix', 'target-port-range', 'target-
  fqdn', or 'target-uri'), the DOTS server appends the parameter values
  in 'alias-name' with the corresponding parameter values in 'target-
  prefix', 'target-port-range', 'target-fqdn', or 'target-uri'.

  The DOTS server indicates the result of processing the PUT request
  using CoAP response codes.  CoAP 2.xx codes are success.  CoAP 4.xx
  codes are some sort of invalid requests (client errors).  COAP 5.xx
  codes are returned if the DOTS server is in an error state or is
  currently unavailable to provide mitigation in response to the
  mitigation request from the DOTS client.

  Figure 10 shows an example response to a PUT request that is
  successfully processed by a DOTS server (i.e., CoAP 2.xx response
  codes).  This version of the specification forbids 'cuid' and 'cdid'
  (if used) to be returned in a response message body.

  {
    "ietf-dots-signal-channel:mitigation-scope": {
        "scope": [
          {
            "mid": 123,
            "lifetime": 3600
          }
        ]
      }
  }

                      Figure 10: 2.xx Response Body

  If the request is missing a mandatory attribute, does not include
  'cuid' or 'mid' Uri-Path options, includes multiple 'scope'
  parameters, or contains invalid or unknown parameters, the DOTS
  server MUST reply with 4.00 (Bad Request).  DOTS agents can safely
  ignore comprehension-optional parameters they don't understand.

  A DOTS server that receives a mitigation request with a lifetime set
  to '0' MUST reply with a 4.00 (Bad Request).

  If the DOTS server does not find the 'mid' parameter value conveyed
  in the PUT request in its configuration data, it MAY accept the
  mitigation request by sending back a 2.01 (Created) response to the
  DOTS client; the DOTS server will consequently try to mitigate the
  attack.  A DOTS server could reject mitigation requests when it is

Reddy, et al.            Expires August 2, 2019                [Page 22]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  near capacity or needs to rate-limit a particular client, for
  example.

  If the DOTS server finds the 'mid' parameter value conveyed in the
  PUT request in its configuration data bound to that DOTS client, it
  MAY update the mitigation request, and a 2.04 (Changed) response is
  returned to indicate a successful update of the mitigation request.

  The relative order of two mitigation requests, having the same
  'trigger-mitigation' type, from a DOTS client is determined by
  comparing their respective 'mid' values.  If two mitigation requests
  with the same 'trigger-mitigation' type have overlapping mitigation
  scopes, the mitigation request with the highest numeric 'mid' value
  will override the other mitigation request.  Two mitigation requests
  from a DOTS client have overlapping scopes if there is a common IP
  address, IP prefix, FQDN, URI, or alias-name.  To avoid maintaining a
  long list of overlapping mitigation requests (i.e., requests with the
  same 'trigger-mitigation' type and overlapping scopes) from a DOTS
  client and avoid error-prone provisioning of mitigation requests from
  a DOTS client, the overlapped lower numeric 'mid' MUST be
  automatically deleted and no longer available at the DOTS server.
  For example, if the DOTS server receives a mitigation request which
  overlaps with an existing mitigation with a higher numeric 'mid', the
  DOTS server rejects the request by returning 4.09 (Conflict) to the
  DOTS client.  The response includes enough information for a DOTS
  client to recognize the source of the conflict as described below in
  the 'conflict-information' subtree with only the relevant nodes
  listed:

  conflict-information:  Indicates that a mitigation request is
      conflicting with another mitigation request.  This optional
      attribute has the following structure:

      conflict-cause:  Indicates the cause of the conflict.  The
        following values are defined:

        1:  Overlapping targets. 'conflict-scope' provides more details
            about the conflicting target clauses.

      conflict-scope:  Characterizes the exact conflict scope.  It may
        include a list of IP addresses, a list of prefixes, a list of
        port numbers, a list of target protocols, a list of FQDNs, a
        list of URIs, a list of alias-names, or a 'mid'.

  If the DOTS server receives a mitigation request which overlaps with
  an active mitigation request, but both having distinct 'trigger-
  mitigation' types, the DOTS server MUST deactivate (absent explicit
  policy/configuration otherwise) the mitigation request with 'trigger-

Reddy, et al.            Expires August 2, 2019                [Page 23]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  mitigation' set to false.  Particularly, if the mitigation request
  with 'trigger-mitigation' set to false is active, the DOTS server
  withdraws the mitigation request (i.e., status code is set to '7' as
  defined in Table 2) and transitions the status of the mitigation
  request to '8'.

  Upon DOTS signal channel session loss with a peer DOTS client, the
  DOTS server MUST withdraw (absent explicit policy/configuration
  otherwise) any active mitigation requests overlapping with mitigation
  requests having 'trigger-mitigation' set to false from that DOTS
  client, as the loss of the session implictily activates these
  preconfigured mitigation requests and they take precedence.  Note
  that active-but-terminating period is not observed for mitigations
  withdrawn at the initiative of the DOTS server.

  DOTS clients may adopt various strategies for setting the scopes of
  immediate and pre-configured mitigation requests to avoid potential
  conflicts.  For example, a DOTS client may tweak pre-configured
  scopes so that the scope of any overlapping immediate mitigation
  request will be a subset of the pre-configured scopes.  Also, if an
  immediate mitigation request overlaps with any of the pre-configured
  scopes, the DOTS client sets the scope of the overlapping immediate
  mitigation request to be a subset of the pre-configured scopes, so as
  to get a broad mitigation when the DOTS signal channel collapses and
  maximize the chance of recovery.

  If the request is conflicting with an existing mitigation request
  from a different DOTS client, the DOTS server may return 2.01
  (Created) or 4.09 (Conflict) to the requesting DOTS client.  If the
  DOTS server decides to maintain the new mitigation request, the DOTS
  server returns 2.01 (Created) to the requesting DOTS client.  If the
  DOTS server decides to reject the new mitigation request, the DOTS
  server returns 4.09 (Conflict) to the requesting DOTS client.  For
  both 2.01 (Created) and 4.09 (Conflict) responses, the response
  includes enough information for a DOTS client to recognize the source
  of the conflict as described below:

  conflict-information:  Indicates that a mitigation request is
      conflicting with another mitigation request(s) from other DOTS
      client(s).  This optional attribute has the following structure:

      conflict-status:  Indicates the status of a conflicting mitigation
        request.  The following values are defined:

        1:  DOTS server has detected conflicting mitigation requests
            from different DOTS clients.  This mitigation request is
            currently inactive until the conflicts are resolved.
            Another mitigation request is active.

Reddy, et al.            Expires August 2, 2019                [Page 24]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

        2:  DOTS server has detected conflicting mitigation requests
            from different DOTS clients.  This mitigation request is
            currently active.

        3:  DOTS server has detected conflicting mitigation requests
            from different DOTS clients.  All conflicting mitigation
            requests are inactive.

      conflict-cause:  Indicates the cause of the conflict.  The
        following values are defined:

        1:  Overlapping targets. 'conflict-scope' provides more details
            about the conflicting target clauses.

        2:  Conflicts with an existing accept-list.  This code is
            returned when the DDoS mitigation detects source addresses/
            prefixes in the accept-listed ACLs are attacking the
            target.

        3:  CUID Collision.  This code is returned when a DOTS client
            uses a 'cuid' that is already used by another DOTS client.
            This code is an indication that the request has been
            rejected and a new request with a new 'cuid' is to be re-
            sent by the DOTS client.  Note that 'conflict-status',
            'conflict-scope', and 'retry-timer' MUST NOT be returned in
            the error response.

      conflict-scope:  Characterizes the exact conflict scope.  It may
        include a list of IP addresses, a list of prefixes, a list of
        port numbers, a list of target protocols, a list of FQDNs, a
        list of URIs, a list of alias-names, or references to
        conflicting ACLs (by an 'acl-name', typically
        [I-D.ietf-dots-data-channel]).

      retry-timer:  Indicates, in seconds, the time after which the DOTS
        client may re-issue the same request.  The DOTS server returns
        'retry-timer' only to DOTS client(s) for which a mitigation
        request is deactivated.  Any retransmission of the same
        mitigation request before the expiry of this timer is likely to
        be rejected by the DOTS server for the same reasons.

        The retry-timer SHOULD be equal to the lifetime of the active
        mitigation request resulting in the deactivation of the
        conflicting mitigation request.  The lifetime of the
        deactivated mitigation request will be updated to (retry-timer
        + 45 seconds), so the DOTS client can refresh the deactivated
        mitigation request after retry-timer seconds before expiry of
        lifetime and check if the conflict is resolved.

Reddy, et al.            Expires August 2, 2019                [Page 25]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  As an active attack evolves, DOTS clients can adjust the scope of
  requested mitigation as necessary, by refining the scope of resources
  requiring mitigation.  This can be achieved by sending a PUT request
  with a new 'mid' value that will override the existing one with
  overlapping mitigation scopes.

  For a mitigation request to continue beyond the initial negotiated
  lifetime, the DOTS client has to refresh the current mitigation
  request by sending a new PUT request.  This PUT request MUST use the
  same 'mid' value, and MUST repeat all the other parameters as sent in
  the original mitigation request apart from a possible change to the
  lifetime parameter value.

4.4.2.  Retrieve Information Related to a Mitigation

  A GET request is used by a DOTS client to retrieve information
  (including status) of DOTS mitigations from a DOTS server.

  'cuid' is a mandatory Uri-Path parameter for GET requests.

  Uri-Path parameters with empty values MUST NOT be present in a
  request.

  The same considerations for manipulating 'cdid' parameter by server-
  domain DOTS gateways specified in Section 4.4.1 MUST be followed for
  GET requests.

  The 'c' (content) parameter and its permitted values defined in
  [I-D.ietf-core-comi] can be used to retrieve non-configuration data
  (attack mitigation status), configuration data, or both.  The DOTS
  server MAY support this optional filtering capability.  It can safely
  ignore it if not supported.  If the DOTS client supports the optional
  filtering capability, it SHOULD use "c=n" query (to get back only the
  dynamically changing data) or "c=c" query (to get back the static
  configuration values) when the DDoS attack is active to limit the
  size of the response.

  The DOTS client can use Block-wise transfer [RFC7959] to get the list
  of all its mitigations maintained by a DOTS server, it can send
  Block2 Option in a GET request with NUM = 0 to aid in limiting the
  size of the response.  If the representation of all the active
  mitigation requests associated with the DOTS client does not fit
  within a single datagram, the DOTS server MUST use the Block2 Option
  with NUM = 0 in the GET response.  The Size2 Option may be conveyed
  in the response to indicate the total size of the resource
  representation.  The DOTS client retrieves the rest of the
  representation by sending additional GET requests with Block2 Options
  containing NUM values greater than zero.  The DOTS client MUST adhere

Reddy, et al.            Expires August 2, 2019                [Page 26]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  to the block size preferences indicated by the DOTS server in the
  response.  If the DOTS server uses the Block2 Option in the GET
  response and the response is for a dynamically changing resource
  (e.g. "c=n" or "c=a" query), the DOTS server MUST include the ETag
  Option in the response.  The DOTS client MUST include the same ETag
  value in subsequent GET requests to retrieve the rest of the
  representation.

  The following examples illustrate how a DOTS client retrieves active
  mitigation requests from a DOTS server.  In particular:

  o  Figure 11 shows the example of a GET request to retrieve all DOTS
      mitigation requests signaled by a DOTS client.

  o  Figure 12 shows the example of a GET request to retrieve a
      specific DOTS mitigation request signaled by a DOTS client.  The
      configuration data to be reported in the response is formatted in
      the same order as was processed by the DOTS server in the original
      mitigation request.

  These two examples assume the default of "c=a"; that is, the DOTS
  client asks for all data to be reported by the DOTS server.

    Header: GET (Code=0.01)
    Uri-Path: ".well-known"
    Uri-Path: "dots"
    Uri-Path: "mitigate"
    Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
    Observe: 0

          Figure 11: GET to Retrieve all DOTS Mitigation Requests

    Header: GET (Code=0.01)
    Uri-Path: ".well-known"
    Uri-Path: "dots"
    Uri-Path: "mitigate"
    Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
    Uri-Path: "mid=12332"
    Observe: 0

      Figure 12: GET to Retrieve a Specific DOTS Mitigation Request

  If the DOTS server does not find the 'mid' Uri-Path value conveyed in
  the GET request in its configuration data for the requesting DOTS
  client, it MUST respond with a 4.04 (Not Found) error response code.
  Likewise, the same error MUST be returned as a response to a request
  to retrieve all mitigation records (i.e., 'mid' Uri-Path is not
  defined) of a given DOTS client if the DOTS server does not find any

Reddy, et al.            Expires August 2, 2019                [Page 27]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  mitigation record for that DOTS client.  As a reminder, a DOTS client
  is identified by its identity (e.g., client certificate, 'cuid') and
  optionally the 'cdid'.

  Figure 13 shows a response example of all active mitigation requests
  associated with the DOTS client as maintained by the DOTS server.
  The response indicates the mitigation status of each mitigation
  request.

Reddy, et al.            Expires August 2, 2019                [Page 28]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  {
    "ietf-dots-signal-channel:mitigation-scope": {
      "scope": [
        {
          "mid": 12332,
          "mitigation-start": "1507818434",
          "target-prefix": [
                &[Ballot discuss]
The Discuss is not for the editors, but for some ADs ... I'll clear immediately after someone clubs me with a clue-by-four.

Dear APP and SEC ADs, in 1.  Introduction, there's this text:

  EST specifies how to transfer messages securely via HTTP over TLS
  (HTTPS) [RFC2818], where the HTTP headers and media types are used in
  conjunction with TLS.  HTTPS operates over TCP; this document does
  not specify EST over Datagram Transport Layer Security/User Datagram
  Protocol (DTLS/UDP). 

I wonder if the disclaimer about not specifying EST over DTLS/UDP is the right thing to say (it's not wrong, although I'm not sure why the parallel to EST/HTTP/TLS/TCP is EST/DTLS/UDP - isn't there a layer missing?).

Given that we already have CORE, which can be fairly characterized as an HTTP variant running over UDP, and given that we approved an IETF 87 BOF on "DTLS In Constrained Environments (DICE)" yesterday, EST over CORE/DTLS/UDP would be an reasonable proposal to think about in the not-too-distant future (not in the context of this document, of course!)

I know we're not going to speculate about possible future work that hasn't even been chartered yet in an RFC that lasts Internet-Forever, but if EST over DTLS/UDP is going to be mentioned at all, is there anything else that would be helpful to say?

At a minumum, I found myself wondering if "EST over DTLS/UDP" was excluded because of some deep security challenge that would prevent EST from ever working over CORE/DTLS/UDP, or just "silly rabbit, EST uses HTTP, and HTTP doesn't run over DTLS".

Even if there are obvious dragons in that direction, that may not be something to mention in the document, but if it is, I should probably ask now ...

[Ballot discuss]
The Discuss is not for the editors, but for some ADs ... I'll clear immediately after someone clubs me with a clue-by-four.

Dear APP and SEC ADs, in 1.  Introduction, there's this text:

  EST specifies how to transfer messages securely via HTTP over TLS
  (HTTPS) [RFC2818gt;                              |
        |  Token: 0x4a                            | Registration
        |  Observe: 0                              |
        +----------------------------------------->|
        |                                          |
        |  2.05 Content                            |
        |  Token: 0x4a                            | Notification of
        |  Observe: 12                            | the current state
        |  status: "attack-mitigation-in-progress" |
        |                                          |
        |<-----------------------------------------+
        |  2.05 Content                            |
        |  Token: 0x4a                            | Notification upon
        |  Observe: 44                            | a state change
        |  status: "attack-successfully-mitigated" |
        |                                          |
        |<-----------------------------------------+
        |  2.05 Content                            |
        |  Token: 0x4a                            | Notification upon
        |  Observe: 60                            | a state change
        |  status: "attack-stopped"                |
        |<-----------------------------------------+
        |                                          |
                            ...

          Figure 14: Notifications of Attack Mitigation Status

4.4.2.2.  DOTS Clients Polling for Mitigation Status

  The DOTS client can send the GET request at frequent intervals
  without the Observe Option to retrieve the configuration data of the
  mitigation request and non-configuration data (i.e., the attack
  status).  The frequency of polling the DOTS server to get the
  mitigation status SHOULD follow the transmission guidelines in
  Section 3.1.3 of [RFC8085].

  If the DOTS server has been able to mitigate the attack and the
  attack has stopped, the DOTS server indicates as such in the status.
  In such case, the DOTS client recalls the mitigation request by
  issuing a DELETE request for this mitigation request (Section 4.4.4).

  A DOTS client SHOULD react to the status of the attack as per the
  information sent by the DOTS server rather than performing its own
  detection that the attack has been mitigated.  This ensures that the

Reddy, et al.            Expires August 2, 2019                [Page 34]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  DOTS client does not recall a mitigation request prematurely because
  it is possible that the DOTS client does not sense the DDoS attack on
  its resources, but the DOTS server could be actively mitigating the
  attack because the attack is not completely averted.

4.4.3.  Efficacy Update from DOTS Clients

  While DDoS mitigation is in progress, due to the likelihood of packet
  loss, a DOTS client MAY periodically transmit DOTS mitigation
  efficacy updates to the relevant DOTS server.  A PUT request is used
  to convey the mitigation efficacy update to the DOTS server.  This
  PUT request is treated as a refresh of the current mitigation.

  The PUT request used for efficacy update MUST include all the
  parameters used in the PUT request to carry the DOTS mitigation
  request (Section 4.4.1) unchanged apart from the 'lifetime' parameter
  value.  If this is not the case, the DOTS server MUST reject the
  request with a 4.00 (Bad Request).

  The If-Match Option (Section 5.10.8.1 of [RFC7252]) with an empty
  value is used to make the PUT request conditional on the current
  existence of the mitigation request.  If UDP is used as transport,
  CoAP requests may arrive out-of-order.  For example, the DOTS client
  may send a PUT request to convey an efficacy update to the DOTS
  server followed by a DELETE request to withdraw the mitigation
  request, but the DELETE request arrives at the DOTS server before the
  PUT request.  To handle out-of-order delivery of requests, if an If-
  Match Option is present in the PUT request and the 'mid' in the
  request matches a mitigation request from that DOTS client, the
  request is processed by the DOTS server.  If no match is found, the
  PUT request is silently ignored by the DOTS server.

  An example of an efficacy update message, which includes an If-Match
  Option with an empty value, is depicted in Figure 15.

Reddy, et al.            Expires August 2, 2019                [Page 35]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      Header: PUT (Code=0.03)
      Uri-Path: ".well-known"
      Uri-Path: "dots"
      Uri-Path: "mitigate"
      Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
      Uri-Path: "mid=123"
      Content-Format: "application/dots+cbor"
      If-Match:
      {
      "ietf-dots-signal-channel:mitigation-scope": {
        "scope": [
          {
            "target-prefix": [
                "2001:db8:6401::1/128",
                "2001:db8:6401::2/128"
              ],
            "target-port-range": [
              {
                "lower-port": 80
              },
              {
                "lower-port": 443
              },
              {
                  "lower-port": 8080
              }
            ],
            "target-protocol": [
                6
            ],
            "attack-status": "under-attack"
          }
        ]
      }
      }

                Figure 15: An Example of Efficacy Update

  The 'attack-status' parameter is a mandatory attribute when
  performing an efficacy update.  The various possible values contained
  in the 'attack-status' parameter are described in Table 3.

Reddy, et al.            Expires August 2, 2019                [Page 36]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  +-----------+-------------------------------------------------------+
  | Parameter | Description                                          |
  |    value |                                                      |
  +-----------+-------------------------------------------------------+
  |        1 | The DOTS client determines that it is still under    |
  |          | attack.                                              |
  +-----------+-------------------------------------------------------+
  |        2 | The DOTS client determines that the attack is        |
  |          | successfully mitigated (e.g., attack traffic is not  |
  |          | seen).                                                |
  +-----------+-------------------------------------------------------+

              Table 3: Values of 'attack-status' Parameter

  The DOTS server indicates the result of processing a PUT request
  using CoAP response codes.  The response code 2.04 (Changed) is
  returned if the DOTS server has accepted the mitigation efficacy
  update.  The error response code 5.03 (Service Unavailable) is
  returned if the DOTS server has erred or is incapable of performing
  the mitigation.  As specified in [RFC7252], 5.03 uses Max-Age option
  to indicate the number of seconds after which to retry.

4.4.4.  Withdraw a Mitigation

  DELETE requests are used to withdraw DOTS mitigation requests from
  DOTS servers (Figure 16).

  'cuid' and 'mid' are mandatory Uri-Path parameters for DELETE
  requests.

  The same considerations for manipulating 'cdid' parameter by DOTS
  gateways, as specified in Section 4.4.1, MUST be followed for DELETE
  requests.  Uri-Path parameters with empty values MUST NOT be present
  in a request.

    Header: DELETE (Code=0.04)
    Uri-Path: ".well-known"
    Uri-Path: "dots"
    Uri-Path: "mitigate"
    Uri-Path: "cuid=dz6pHjaADkaFTbjr0JGBpw"
    Uri-Path: "mid=123"

                  Figure 16: Withdraw a DOTS Mitigation

  If the DELETE request does not include 'cuid' and 'mid' parameters,
  the DOTS server MUST reply with a 4.00 (Bad Request).

Reddy, et al.            Expires August 2, 2019                [Page 37]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

  Once the request is validated, the DOTS server immediately
  acknowledges a DOTS client's request to withdraw the DOTS signal
  using 2.02 (Deleted) response code with no response payload.  A 2.02
  (Deleted) Response Code is returned even if the 'mid' parameter value
  conveyed in the DELETE request does not exist in its configuration
  data before the request.

  If the DOTS server finds the 'mid' parameter value conveyed in the
  DELETE request in its configuration data for the DOTS client, then to
  protect against route or DNS flapping caused by a DOTS client rapidly
  removing a mitigation, and to dampen the effect of oscillating
  attacks, the DOTS server MAY allow mitigation to continue for a
  limited period after acknowledging a DOTS client's withdrawal of a
  mitigation request.  During this period, the DOTS server status
  messages SHOULD indicate that mitigation is active but terminating
  (Section 4.4.2).

  The initial active-but-terminating period SHOULD be sufficiently long
  to absorb latency incurred by route propagation.  The active-but-
  terminating period SHOULD be set by default to 120 seconds.  If the
  client requests mitigation again before the initial active-but-
  terminating period elapses, the DOTS server MAY exponentially
  increase (the exponent is 2) the active-but-terminating period up to
  a maximum of 300 seconds (5 minutes).

  Once the active-but-terminating period elapses, the DOTS server MUST
  treat the mitigation as terminated, as the DOTS client is no longer
  responsible for the mitigation.  For example, if there is a financial
  relationship between the DOTS client and server domains, the DOTS
  client stops incurring cost at this point.

  If a mitigation is triggered due to a signal channel loss, the DOTS
  server relies upon normal triggers to stop that mitigation
  (typically, receipt of a valid DELETE request, expiry of the
  mitigation lifetime, or scrubbing the traffic to the attack target).
  In particular, the DOTS server MUST NOT consider the signal channel
  recovery as a trigger to stop the mitigation.

4.5.  DOTS Signal Channel Session Configuration

  A DOTS client can negotiate, configure, and retrieve the DOTS signal
  channel session behavior with its DOTS peers.  The DOTS signal
  channel can be used, for example, to configure the following:

  a.  Heartbeat interval (heartbeat-interval): DOTS agents regularly
      send heartbeats (CoAP Ping/Pong) to each other after mutual
      authentication is successfully completed in order to keep the
      DOTS signal channel open.  Heartbeat messages are exchanged

Reddy, et al.            Expires August 2, 2019                [Page 38]
Internet-Draft        DOTS Signal Channel Protocol          January 2019

      between DOTS agents every 'heartbeat-interval' seconds to detect
      the current status of the DOTS signal channel session.

  b.  Missing heartbeats allowed (missing-hb-allowed): This variable
      indicates the maximum number of consecutive heartbeat messages
      for which a DOTS agent did not receive a response before
      concluding that the session is disconnected or defunct.

  c.  Acceptable signal loss ratio: Maximum retransmissions,
      retransmission timeout value, and other message transmission
      parameters for the DOTS signal channel.

  The same or distinct configuration sets may be used during times when
  a mitigation is active ('mitigating-config') and when no mitigation
  is active ('idle-config').  This is particularly useful for DOTS
  servers that might want to reduce heartbeat frequency or cease
  heartbeat exchanges when an active DOTS client has not requested
  mitigation.  If distinct configurations are used, DOTS agents MUST
  follow the appropriate configuration set as a function of the
  mitigation activity (e.g., if no mitigation request is active, 'idle-
  config'-related values must be followed).  Additionally, DOTS agents
  MUST automatically switch to the other configuration upon a change in
  the mitigation activity (e.g., if an attack mitigation is launched
  after a peacetime, the DOTS agent switches from 'idle-config' to
  'mitigating-config'-related values).

  CoAP Requests and responses are indicated for reliable delivery by
  marking them as Confirmable messages.  DOTS signal channel session
  configuration requests and responses are marked as Confirmable
  messages.  As explained in Section 2.1 of [RFC7252], a Confirmable
  message is retransmitted using a default timeout and exponential
  back-off between retransmissions, until the DOTS server sends an
  Acknowledgement message (ACK) with the same Message ID conveyed from
  the DOTS client.

  Message transmission parameters are defined in Section 4.8 of
  [RFC7252].  The DOTS server can either piggyback the response in the
  acknowledgement message or, if the DOTS server cannot respond
  immediately to a request carried in a Confirmable message, it simply
  responds with an Empty Acknowledgement message so that the DOTS
  client can stop retransmitting the request.  Empty Acknowledgement
  messages are explained in Section 2.2 of [RFC7252].  When the
  response is ready, the server sends it in a new Confirmable message
  which in turn needs to be acknowledged by the DOTS client (see
  Sections 5.2.1 and 5.2.2 of [RFC7252]).  Requests and responses
  exchanged between DOTS agents during peacetime are marked as
  Confirmable messages.

Reddy, et al.            Expires August 2, 2019                [Page 39]
], where the HTTP headers and media types are used in
  conjunction with TLS.  HTTPS operates over TCP; this document does
  not specify EST over Datagram Transport Layer Security/User Datagram
  Protocol (DTLS/UDP). 

I wonder if the disclaimer about not specifying EST over DTLS/UDP is the right thing to say (it's not wrong!).

Given that we already have CORE, which can be fairly characterized as an HTTP variant over UDP, an EST over CORE/DTLS/UDP would be an reasonable question to think about (not in this document, of course!)

If this topic is going to be mentioned at all, is there anything else that would be helpful to say? At a minumum, I found myself wondering if "EST over DTLS/UDP" was excluded because of some deep security challenge that would prevent EST from ever working over CORE/DTLS/UDP, or just "silly rabbit, EST is using HTTP, and HTTP doesn't run over DTLS".

That may not be something to mention in the document, but if it is, I should probably ask now ...

[Ballot comment]
These comments are for the editors. Please consider them as your document moves through the process.

In 3.2.  HTTP Layer

  HTTP 1.1 [RFC2616] and above support persistent connections.  As
  described in Section 8.1 of that RFC, persistent connections may be
  used to reduce network and processing load associated with multiple
  HTTP requests.  EST does not require or preclude persistent HTTP
  connections and their use is out of scope of this specification.

If I understand the final sentence correctly, EST works fine whether you use persistent HTTP connections or not. If that's so, and I suspect it is, that's not "out of scope", that's "good news! we've looked at it, and there's no problem". Please consider removing "and their use is out of scope of this specification".

In 4.4.1.  Server-side Key Generation Request

  The server needs to know a priori about the algorithms supported by
  the client.

I don't know whether this requirement is typical or not, but I wonder - is there some way this might be signaled?

http://www.imc.org/ietf-pkix/pkix-oid.asn for further information about PKIX OID registration procedures.

Please add a sentence to the document noting that this registry is not maintained by IANA. It might also be appropriate to move this notice to another section of the document.

Second, in the Well-Known URIs registry located at:

http://www.iana.org/assignments/well-known-uris/well-known-uris.xml

IANA will register the following URI suffix as follows:

URI suffix: est
Change controller: IETF
Reference: [ RFC-to-be ]
Related Information:
Date Registered: [ TBD-at-registration ]
Date Modified:

IANA will submit this request to the registry expert for review.

Third, in the Application Media Types registry located at:

http://www.iana.org/assignments/media-types/application

a new application media type will be registered:

csrattrs

with a reference of [ RFC-to-be ]. The template for the registration is provided in Section 6 of the approved document.

Fourth, the authors note that "The application/pkcs7-mime content type defines the optional "smime-type" parameter [RFC5751]. The smime-type parameter for Server-side Key Generation Response is server-generated-key." The optional parameters for the "smime-type" are not registered separately. Instead, they are provided as part of the reference to the application/pkcs7-mime registration.

IANA Question -> Would the authors like to add a reference to this document to the established [RFC5751] reference for the application/pkcs7-mime media type?

IANA understands that these actions are the only ones required upon approval of this document.

Note:  The actions requested in this document will not be completed until the document has been approved for publication as an RFC. This message is only to confirm what actions will be performed.

The following Last Call announcement was sent out:

From: The IESG
To: IETF-Announce
CC:
Reply-To: ietf@ietf.org
Sender:
Subject: Last Call:  (Enrollment over Secure Transport) to Proposed Standard


The IESG has received a request from the Public-Key Infrastructure
(X.509) WG (pkix) to consider the following document:
- 'Enrollment over Secure Transport'
  as Proposed Standard

The IESG plans to make a decision in the next few weeks, and solicits
final comments on this action. Please send substantive comments to the
ietf@ietf.org mailing lists by 2013-06-24. Exceptionally, comments may be
sent to iesg@ietf.org instead. In either case, please retain the
beginning of the Subject line to allow automated sorting.

Abstract


  This document profiles certificate enrollment for clients using
  Certificate Management over CMS (CMC) messages over a secure
  transport.  This profile, called Enrollment over Secure Transport
  (EST), describes a simple yet functional certificate management
  protocol targeting Public Key Infrastructure (PKI) clients that need
  to acquire client certificates and associated Certification Authority
  (CA) certificate(s).  It also supports client-generated public/
  private key pairs as well as key pairs generated by the CA.




The file can be obtained via
http://datatracker.ietf.org/doc/draft-ietf-pkix-est/

IESG discussion can be tracked via
http://datatracker.ietf.org/doc/draft-ietf-pkix-est/ballot/


No IPR declarations have been submitted directly on this I-D.