Re: [dhcwg] Anyone interested in continuing draft-ietf-dhc-dhcpv6-prefix-pool-opt?

<> Mon, 26 August 2013 06:47 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 38A0211E8155 for <>; Sun, 25 Aug 2013 23:47:36 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -2.247
X-Spam-Status: No, score=-2.247 tagged_above=-999 required=5 tests=[AWL=-0.000, BAYES_00=-2.599, HELO_EQ_FR=0.35, HTML_MESSAGE=0.001, UNPARSEABLE_RELAY=0.001]
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id EHypxQwmzPTJ for <>; Sun, 25 Aug 2013 23:47:29 -0700 (PDT)
Received: from ( []) by (Postfix) with ESMTP id F3E5211E8130 for <>; Sun, 25 Aug 2013 23:47:28 -0700 (PDT)
Received: from (unknown [xx.xx.xx.200]) by (ESMTP service) with ESMTP id 94F241B81E3; Mon, 26 Aug 2013 08:47:27 +0200 (CEST)
Received: from (unknown []) by (ESMTP service) with ESMTP id 7057615805E; Mon, 26 Aug 2013 08:47:27 +0200 (CEST)
Received: from ([]) by ([]) with mapi; Mon, 26 Aug 2013 08:47:27 +0200
From: <>
To: Roberta Maglione <>
Date: Mon, 26 Aug 2013 08:47:24 +0200
Thread-Topic: [dhcwg] Anyone interested in continuing draft-ietf-dhc-dhcpv6-prefix-pool-opt?
Thread-Index: Ac6gMcvnGYEsj4R6SxqJ4VCvnwZaDwB8UWHw
Message-ID: <>
References: <> <> <> <> <> <> <> <>
In-Reply-To: <>
Accept-Language: fr-FR
Content-Language: fr-FR
acceptlanguage: fr-FR
Content-Type: multipart/alternative; boundary="_000_94C682931C08B048B7A8645303FDC9F36EEEE4E959PUEXCB1Bnante_"
MIME-Version: 1.0
X-PMX-Version:, Antispam-Engine:, Antispam-Data: 2013.8.26.54519
Cc: " WG" <>, Ralph Droms <>
Subject: Re: [dhcwg] Anyone interested in continuing draft-ietf-dhc-dhcpv6-prefix-pool-opt?
X-Mailman-Version: 2.1.12
Precedence: list
List-Id: <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Mon, 26 Aug 2013 06:47:36 -0000

Hi Roberta,

Thank you for your answer.

Please see inline.


De : Roberta Maglione []
Envoyé : vendredi 23 août 2013 20:52
Cc : Ralph Droms; WG
Objet : Re: [dhcwg] Anyone interested in continuing draft-ietf-dhc-dhcpv6-prefix-pool-opt?

Hi Med, Leaf,

I fully understand the requirement for network automation versus manual configuration, but I still don't think defining a new dhcp option is the right direction to go in order to do routes aggregations in a router.

[Med] Can you please elaborate why this is not a right direction? Isn't prefix assignment a key point which would impact route aggregation? Doesn't prefix assignment has requirements on routing? Couldn't dhcp be tweaked to optimize prefix assignments and avoid increasing the size of routing tables? Wouldn't be there a need to sync prefix assignment state and the routing controller which would interact with a routing engine? Wouldn't be optimize to get rid of that interface and let dhcp pass that information directly to the routing protocol without passing through a routing controller?

As somebody else already pointed out here, if you need a way to automatically interact with the router i2rs is the right place to discuss your requirements.

[Med] this is not conflicting. I2rs does not define does (aim to) define an interface. If we want to use i2rs terminology, this draft proposes to use dhcp as the i2rs interface for the purpose of tweaking route aggregation, and only for this need because this is the place where prefixes are managed.

In addition you said:

>  I confirm this option is not a per-customer configuration parameter.

 If it is not a per-client configuration why  would you like to use DHCP for that?

[Med] The per-client assigned prefix is part of the prefix returned by this option. There is dependency between routing and prefix assignment.

I understand that in DHCP Server there are already some global parameters like the pool of addresses to be used for the clients, but my understanding of DHCP behavior , and please correct me if I'm wrong, is that the DHCP Server provides a per-client configuration.

How would it work with this new option? How does the DHCP Server know if it has to add or not this option if it is not based on client profile or on client request?

Do you plan to insert the option in all  the DHCP Server replies/to all clients?

[Med] The current text says it is inserted in all messages by relay. The server may not inserted the option in all responses. It can insert the option only if there are changes.

I assume the router is going to install the route once and what does it do if it gets the same routes more the one? Just ignore? It seems like a waste of cycles on the router.

[Med] The server is not requirement to insert the option in all responses. BTW, in any dynamic provisioning protocol there is a provisioning epoch to be followed to ensure consistent policies are enforced. Inserting the option on all relay-agent messages is a way to preserve a provisioning epoch.

In my opinion having global parameters on DHCP Server seems  a major change in the DHCP model.

[Med] There is no change in the model. The prefix returned in the option may be for instance a per-region prefix. See the example in:

   "... In this example, link C is a regional backbone for an ISP.  Link E is
   also a regional backbone for that ISP.  Relays A, B, C and D are PE
   routers, and Links A, B, F and G are actually link aggregators with
   individual layer 2 circuits to each customer-\u002Dfor example, the
   relays might be DSLAMs or cable head-end systems.  At each customer
   site we assume there is a single CPE device attached to the link. ..."

draft-ietf-dhc-dhcpv6-prefix-pool-opt does only return the aggregate for each relays A, B, C and D. As you can see draft-lemon-*, this information (i.e., /40s) is configured to the server.



On Fri, Aug 23, 2013 at 2:15 AM, <<>> wrote:
Hi Roberta,

Yes, as indicated in the document, manual configuration is an option... but it has its limits too.

This proposal is a contribution to add more automation to network configuration without requiring an additional dynamic protocol to drive how aggregates are built in a router co-located with a requestor, and therefore interact in a more dynamic fashion with a routing protocol (e.g., drive route withdrawals, etc.).

Of course, some routers can offer some features to optimize the size of routing tables and prevent from injecting (very) specific entries. But still this behavior is implementation-specific and does not provide the same aggregation level as the one proposed in this document.

Unlike implementation-specific behaviors, this proposal is deterministic since it is fully controlled by the entity which has the full knowledge of prefix related states and network policies: e.g., the server has the knowledge of prefix assignment, prefix assignment policies, prefix aggregates, etc.

I confirm this option is not a per-customer configuration parameter.


De : Roberta Maglione [<>]
Envoyé : jeudi 22 août 2013 22:31
Cc : Ralph Droms;<> WG

Objet : Re: [dhcwg] Anyone interested in continuing draft-ietf-dhc-dhcpv6-prefix-pool-opt?


Maybe I'm missing something here, but I'm struggling to see the value added by this new option in terms of route aggregation functionality.

Today with IPv4 if I need to summarize some routes I manually configure on the router a summary/aggregate route and I announce it into the routing protocol. Moving to IPv6 you could do the same thing, I don't quite get what's wrong with that?

You say you would like to have an automatic way to tell the PE to aggregate the routes, but if I understand correctly the proposal what you are doing here is only moving the configuration of the summary route from the PE to the DHCPv6 Server; what do you really save here?

In addition the route aggregation is not a per customer configuration, it would be per box or per service configuration so why do you want to add it to customers' profile in DCHPv6 Server?



On Thu, Aug 22, 2013 at 9:45 AM, <<>> wrote:

IMHO, draft-ietf-dhc-dhcpv6-agentopt-delegate does not cover the same objectives as in draft-ietf-dhc-dhcpv6-prefix-pool-opt.

draft-ietf-dhc-dhcpv6-prefix-pool-opt aims to provide a dynamic means to trigger route advertisement actions and to control the route aggregates to be injected using a routing protocol. For example, a router can be told by the DHCP server to advertise an aggregate even if not all individual prefixes are assigned to customer located behind that router. This is a measure that can help in optimizing routing tables and avoid injecting very specific routes. Snooping the assignment and then guide the route advertisement actions may not be lead to the same optimized routing tables, because there will be "holes" that will prevent aggregating routes.

Having an explicit channel like the one specified in draft-ietf-dhc-dhcpv6-prefix-pool-opt is superior IMHO.


>-----Message d'origine-----
>De :<> [<>] De la part de
>Ralph Droms
>Envoyé : jeudi 22 août 2013 14:48
>À : Alexandru Petrescu
>Cc :<> WG
>Objet : Re: [dhcwg] Anyone interested in continuing draft-ietf-dhc-dhcpv6-
>On Aug 21, 2013, at 9:24 AM 8/21/13, Alexandru Petrescu
><<>> wrote:
>> Hi,
>> One point I think is essential is the installment of routes in the DHCP
>> Relay upon Prefix Assignment.
>> The base DHCPv6 Prefix Delegation RFC does not stipulate that DHCP must
>> install a route in the DHCP Relay upon delegation.
>> This draft seems to at least assume it, and to describe much more about
>> it: how various parts of assigned prefixes are aggregated and
>> I support it.
>After a quick read, draft-ietf-dhc-dhcpv6-agentopt-delegate seems to have
>been aimed at the same problem.  If I have that right, it might be
>instructive to review the dhc WG mailing list discussion that lead to the
>abandonment of draft-ietf-dhc-dhcpv6-agentopt-delegate.
>- Ralph
>> Alex
>> Le 21/08/2013 14:41,<> a écrit :
>>> Hi Tomek,
>>> I do still think draft-ietf-dhc-dhcpv6-prefix-pool-opt documents a
>>> useful feature in order to have more automation and also control
>>> routes aggregation instead of relying on proprietary behaviors of
>>> each implementation. Of course, part of these objectives can be
>>> achieved if routes are installed manually or use an out of band
>>> mechanism to enforce routing aggregation policies. Still, the
>>> proposal in draft-ietf-dhc-dhcpv6-prefix-pool-opt is superior
>>> because the DHCP server has the knowledge of the prefix assignments;
>>> and therefore routes can be triggered with dhcpv6 .
>>> A way to progress with this document is to target the Experimental
>>> track. Based on the experience that will be gained in real
>>> deployments, the status can be revisited if required.
>>> Cheers, Med
>>>> -----Message d'origine----- De :<>
>>>> [<>] De la part de Tomek Mrugalski
>>>> Envoyé : lundi 19 août 2013 16:52 À : dhcwg Objet : [dhcwg] Anyone
>>>> interested in continuing draft-ietf-dhc-dhcpv6- prefix-pool-opt?
>>>> During Berlin meeting chairs asked if there is still interest in
>>>> the prefix-pool-option. There was nobody interested in the work in
>>>> the room. The unanimous consensus in the room was to drop it. I
>>>> just wanted to confirm that on the list.
>>>> If you are interested in this work, want to support it and
>>>> participate in it, please let us know by replying to the mailing
>>>> list. Otherwise we'll drop this work and mark that draft as a dead
>>>> WG document.
>>>> Please respond within 2 weeks (until Sep. 2nd).
>>>> Bernie & Tomek _______________________________________________
>>>> dhcwg mailing list<>
>>> _______________________________________________ dhcwg mailing list
>> _______________________________________________
>> dhcwg mailing list
>dhcwg mailing list
dhcwg mailing list<>