DHCPv4 Vendor-specific Message
draft-ietf-dhc-dhcpv4-vendor-message-01

I think adding a vendor-specific message to DHCP is a fine thing to do.

However, I have a relatively minor issue with the current text. I don't think it is correct to suggest that a purpose of this is experimentation. Experiments are not supposed to escape into the wider Net and an experimental message would be treated differently from a vendor-specific message.

While it is true that a vendor may choose to experiment under their enterprise number, the fact that it is an experiment is actually hidden from the world, and we don't consider listing all of the other uses to which this message might be put.

I would be happy to see the reference to experimentation removed from the Abstract and the Introduction.

I generally don't object to this document going forward, however I believe there are some minor points that needs to be addressed/discussed first:

1) In Section 3

   Clients and servers MAY chose to support this message; those that do
   not, MUST discard the message.  Relay agents SHOULD relay these
   messages as they would other DHCPv4 messages unless the relay agent
   understands the specific message and knows that the message was
   directed at it.

This sounds as if this document is trying to put requirements on implementations that are not compliant with this document.
Is this section just describes something which follows from DHCPv4
extensibility rules?

2) In Section 4:

   The option-data field MUST be encoded as a sequence of code/length/
   value fields of identical format to the DHCP options field and is
   identical to the option-data field of Vendor-Identifying Vendor
   Options [RFC3925].

This reference looks Normative to me.

   The option codes are defined by the vendor
   identified in the enterprise-number field and are not managed by
   IANA.  Option codes 0 and 255 have no pre-defined interpretation or

Are you talking about subopt-code here?

   format.  Each of the encapsulated options is formatted as follows:

3) In Section 4:

   Clients, relay agents, and/or servers supporting the Vendor Message
   Option MUST support [RFC3396].

This reference looks Normative to me.

4) Taking over Pasi's DISCUSS:

Section 1 says "New message codes are assigned through IETF Standards
Action". According to the registry and RFC 2939, the policy is
currently "IETF Consensus" -- is this just a typo, or is the intent to
change the policy?

Pasi also wrote: According to the WG chair (Ted Lemon, 2010-02-16), the intent was not to change the current policy, so in Section 1, "Standards Action" should be changed to "IETF Consensus" (or perhaps "IETF Review" in 5226 terms).

I view this document as effectively making it so no review is needed of new DHCP options. I don't think theWG charter allows this. I also view this as very bad thing given the goals of the IETF and would likely move to abstain. If the WG wanted to move to something that allowed a large code space with a easier review process, I would have no objects to that.

Section 4 - the document specify that option-len and subopt-len are expressed in octets

Should this document contain (or point to) more guidance that would encourage the use and development of interoperable DHCP options? It is currently missing any pressure to discourage the replication of already standardized behavior and encourage bringing common functionality through the standards process.

  Two concerns were raised in the Gen-ART Review by Elwyn Davies
  that need to be addressed:

  s3, para 3: This paragraph contains weasel words about 'good
  networking practices' that 'MUST be used'.  This is untestable
  as it stands.  Is this anything more than dealing with non-replies,
  not flooding the network with repeats, and not hanging up if the
  partner never does answer?  Also, servers or clients that do not
  (yet) implement this capability do not have a defined behaviour
  when receiving this new type of message.  There is a assumption
  that they will be silently dropped without causing any problems.

  s7 and s4, next to last para: s4 specifies RFC 3396 as 'MUST support'
  This makes RFC 3396 a normative reference, not informative.  Arguably
  the last para of s4 makes RFC 3925 normative as well.