Hi Spencer,
Thanks for the review! See some comments inline.
On Tue, Dec 2, 2008 at 11:18 PM, Spencer Dawkins
<spencer(_at_)wonderhamster(_dot_)org> wrote:
I have been selected as the General Area Review Team (Gen-ART)
reviewer for this draft (for background on Gen-ART, please see
http://www.alvestrand.no/ietf/gen/art/gen-art-FAQ.html).
Please resolve these comments along with any other Last Call comments
you may receive.
Document: draft-korhonen-mip4-service-06
Reviewer: Spencer Dawkins
Review Date: 2008-12-02
IETF LC End Date: 2008-12-24
IESG Telechat date: (not known)
Summary: This draft is on the right track for publication as Informational
(should it be standards-track, if you're expecting this to be widely
deployed? But I'll leave that to the IESG).
I would have nothing against standards track RFC.
I do have comments, especially involving 2119 language.
Comments:
(Can Jouni's e-mail address really be
Email: jouni(_dot_)nospam(_at_)gmail(_dot_)com
?)
Yes, this is my valid gmail e-mail address ;)
1. Introduction
This document describes a Service Selection Extension for Mobile IPv4
that is intended to assist home agents to make specific service
selections for the mobility service subscription during the
registration procedure. The service selection may affect home agent
routing decisions, Home Address assignment policies, firewall
settings, and security policies. The Service Selection extension
SHOULD be used in every Registration Request that makes a new
registration to the home agent. The Service Selection extension from
the Registration Request MAY be echoed back in the Registration
Reply.
Spencer: I don't usually see 2119 normative language in the introduction of
Internet Drafts... at a minimum, these statements appear before the
requirements key words are introduced in section 2. I THINK most of these
requirements are restated later in the document anyway, so they could
probably be dropped here.
Ok, good point. I'll will remove the 2119 language from the Introduction.
In absence of a specifically indicated service the home agent MUST
act as if the default service, plain Internet access had been
requested. There is no absolute requirement that this default
service be allowed to all subscribers, but it is highly RECOMMENDED
in order to avoid having normal subscribers employ operator-specific
configuration values in order to get basic service.
Some of the potential use-cases were listed earlier in this section.
The general aim is better manageability of services and service
provisioning from both operators and service providers point of view.
However, it should be understood that there are potential deployment
possibilities where selecting a certain service may restrict
simultaneous access to other services from an user point of view.
For example, services may be located in different administrative
domains or external customer networks that practice excessive
filtering of inbound and outbound traffic.
Spencer: I wasn't clear on who this understanding is directed to - it almost
reads like you're warning users that bad things might happen if you use a
specific service, but surely the user specifies the service because an
operator requires this?
We had this discussion earlier on RFC5149.. and concerns were raised
whether the Service Selection is a potential tool for enabling walled
gardens etc. Thus this note here was added to emphasize that point.
3. Service Selection Extension
At most one Service Selection extension MAY be included in any Mobile
IPv4 Registration Request message. It SHOULD be included at least in
Spencer: seems to be missing a qualifier: "When a non-default service is
selected, the Service Selection extension SHOULD be included ..."?
Spencer: If this is qualified, could the SHOULD be a MUST?
Hmm.. right. New text would be:
At most one Service Selection extension MAY be included in any Mobile
IPv4 Registration Request message. When a non-default service is selected,
the Service Selection extension SHOULD be included at least in
the Registration Request message that is sent for the initial binding
registration when the mobile node and the home agent do not have an
existing binding. The Service Selection extension MUST be placed in
the Registration Request message as follows:
Spencer: If it remains as SHOULD, what happens if the Service Selection
extension is not included in the initial binding registration, but is
included in subsequent binding registrations?
The first RRQ would be treated as the selection of the "default
service". The subsequent RRQs with the service selection would cause
reauthorization & evaluation of the requested service. If the newly
requested service conflicts with the "default service" from the HA
point of view, then an appropriate error is returned and the binding
is dropped.
the Registration Request message that is sent for the initial binding
registration when the mobile node and the home agent do not have an
existing binding. The Service Selection extension MUST be placed in
the Registration Request message as follows:
o When present the extension MUST appear after the MN-NAI extension,
if the MN-NAI is also present in the message
o If the extension was added by the mobile node to a Registration
Request it MUST appear prior any authentication-enabling
extensions [RFC3344][RFC4721]
Spencer (editorial): s/prior/before/ or s/prior/prior to/
Ack.
o In the event the foreign agent adds the Service Selection
extension to a Registration Request, the extension MUST appear
prior to any Foreign-Home authentication-enabling extensions
[RFC3344]
4.1. Mobile Node Considerations
A mobile node or its proxy representative MAY include the Service
Selection extension into any Registration Request message. The
Service Selection extension can be used with any mobile node
identification method. The extension is used to identify the service
to be associated with the mobility session and SHOULD only be
Spencer: this seems to be more restrictive than previous text that allowed
the extension to be included in non-initial registration request messages...
Hmm.. earlier text said "..the Service Selection extension SHOULD be
included at least in the Registration Request message that is sent for
the initial binding registration.." I find these equally restricting.
included into the initial Registration Request message sent to a home
agent. If the mobile node wishes to change the selected service, it
is RECOMMENDED that the mobile node de-register the existing binding
Spencer: why RECOMMENDED and not REQUIRED? RECOMMENDED means that home agents
must handle the case where the SHOULD isn't observed anyway.
We wanted to leave the possibility of MNs changing the selected
service in fly without tearing down the binding (which would mean
break in the IP connectivity from MIP point of view). Depending on the
policies on the HA and the selected service, the change of the
"selected service" may not cause a change of the HoA. Since this
behavior is tied to deployment time policies and requires out of band
knowledge of the services on the MN, the general recommendation is to
bring the existing service down first.
with the home agent before proceeding with a binding registration for
a different service. The provisioning of the service identifiers to
the mobile node or its proxy representative is out of scope of this
specification.
6. IANA Considerations
A new Mobile IPv4 skippable Extension type is required for the
following new Extension described in Section 3. The Extension type
must be from the 'skippable Extension' range (128-255):
Service Selection Extension is set to TBD
A new Mobile IPv4 registration denied by home agent error code is
required. The error code must be allocated from the 'Error Codes
from the Home Agent' range (128-192):
SERVICE_AUTHORIZATION_FAILED is set to TBD
Spencer: (if you revise this draft, you probably want to use "TBD1", "TBD2",
etc. so that it's clearer to IANA which "TBD" gets replaced with each
allocated value)
Good point. I'll fix that.
Cheers,
Jouni
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf