[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: New (-02) version of IPv6 CPE Router draft is available for review



Perhaps we could have a more constructive conversation about solutions if we had a detailed description of the requirements for subscriber network and device operation.

Certainly assigning both ULA and GUA prefixes across the subscriber network is doable, at the cost of extra complexity and the assumption that RFC 3484 will actually . Use of ULA will give a stable address - but that stable address is only useful in the unlikely case the home network operator actually enters the IPv6 address for the gateway instead of using DNS or other naming.

And, it's up to the ISP to provide global prefix stability...

- Ralph

On Jul 21, 2008, at Jul 21, 2008,11:57 AM, Hemant Singh (shemant) wrote:

I repeat myself. I am only discussing the case when the CPE Router is powered up without connecting the router WAN interface to the SP. This is also the first time the CPE Router is powered up. Further this is NOT a corner case. Moreover, all IPv4 home router customers are used to one stable private IPv4 address to access their say, Linksys router, no matter when and what global IPv4 address gets assigned to the WAN interface. Why not use one ULA for the same purpose in IPv6? With your proposal, if the WAN GUA is used to configure the CPE Router, that GUA may change over time and now the user has to remember a new GUA.

Sorry, I don't think this user experience should be brushed away for IPv6.

Hemant

-----Original Message-----
From: Ralph Droms (rdroms)
Sent: Monday, July 21, 2008 11:50 AM
To: Hemant Singh (shemant)
Cc: Ralph Droms (rdroms); Rémi Denis-Courmont; Alain Durand; v6ops@ops.ietf.org ; Wes Beebee (wbeebee) Subject: Re: New (-02) version of IPv6 CPE Router draft is available for review

Sorry; yes, please s/GUA/ULA>

The discussion is not irrelevant.  I've seen references in the thread
to generically enabling the LAN interface before the WAN interface.
One of the points I explained in my e-mail is that *once the CPE has received a prefix* the order of interface enabling is irrelevant.

I also explicitly discussed the corner case you describe. I don't understand why we want to add complexity to allow the customer to configure the CPE router before connecting to the WAN. Once you understand that the CPE router can retain the delegated prefix, there is no point to worrying about ULA and then renumbering for global addresses. Simply require that the CPE router connect to the WAN long enough to get a delegated prefix and then allow the subscriber to do any customized config.

- RAlph

On Jul 21, 2008, at Jul 21, 2008,11:39 AM, Hemant Singh (shemant) wrote:


-----Original Message-----
From: Ralph Droms (rdroms)
Sent: Monday, July 21, 2008 11:21 AM
To: Rémi Denis-Courmont
Cc: Ralph Droms (rdroms); Alain Durand; Hemant Singh (shemant);
v6ops@ops.ietf.org ; Wes Beebee (wbeebee)
Subject: Re: New (-02) version of IPv6 CPE Router draft is available
for review

The only time the CPE router needs to use GUA is if it has never
received a delegated prefix.

Huh? Did you mean to say ULA?  Anyhow, the rest of your discussion
does not map to the problem at hand.  Our problem is that I have
bought this standalone CPE router from Best Buy (in the U.S.) and
now I want to configure it at home via the web for the first time.
I have still not connected the router to the SP WAN.  I just connect
an Ethernet cable between the CPE Router LAN interface and my PC and
power up the router.  Now I need to configure the router via the web.
Initially we had suggested to use link-local address of the LAN
interface to access the router from the web. But Remi said that needs
the full network name interface configuration in some browsers (we
didn't need that in our test at Cisco) and also the access has
some non-compliance with HTTP.  So then we said, we'll use a ULA.
At this juncture of the router operation/configuration, the router
doesn't have any GUA.  A lot of folks in private emails to us have
agreed to supporting ULA on the LAN interface(s) of the CPE Router.
We really don't think a coexisting ULA with GUA on the LAN in a
problem for any data forwarding or source address selection.

We have one solution using ULA. If a better solution is available for
the problem described above, we are open to it.

Thanks.

Hemant

Prefix delegation uses the same leasing mechanism as address
assignment; if the WAN interface on the CPE router is not up, the CPE
router can still use > the previously delegated prefix on the
subscriber network.

So, as long as the CPE router has been connected to the ISP network
at least once and has obtained a delegated prefix, there is no need
for GUA. To avoid unnecessary complexity, I would strongly recommend
that the CPE router be required to complete an initial prefix
delegation operation when first
connected to the ISP before enabling any downstream interfaces.
Once that prefix delegation takes place, it is immaterial whether the
LAN or WAN interface comes up first.

This discussion brings up the problem of subscriber network operation
- is it an explicit goal of this document that the subscriber
network be able to operate even if the WAN interface is not up?
What are the expectations > for that unconnected operation and what
services must the CPE router provide to meet those expectations?

According to the spec, the requesting router can only use the
delegated prefix until the lease on the prefix expires.  We hadn't
thought about relaxing > that restriction to allow the requesting
router to continue to use the delegated prefix until it has a WAN
connection back to the delegating router.

- Ralph


On Jul 21, 2008, at Jul 21, 2008,10:57 AM, Rémi Denis-Courmont wrote:

Le vendredi 18 juillet 2008 19:25:54 Alain Durand, vous avez écrit :
Supporting both ULA & GUA at the same time is also a source of
complexity and confusion. The key problem I see is with external
referrals in multi-party communications where some of the hosts are
inside, and some are outside. Mixing ULA & GUA can have complex
consequences, and again generates service call.

Yeah. Broken RFC3484 implementations will do just that. But all nodes
(broken and non-broken RFC3484 implementations alike) will *break*
without ULA, until we have *instantaneous* 100%-reliable and
100%-available upstream connections (which we will NEVER have).
Without this, the network will simply not work until the ISP
connection is established (if ever), which is a total non-starter.
Therefore, it seems like a total non-question that ULA is the way to
go.

Also, if I read the text correctly, if the WAN interface gets
configured first, no ULA are generated. Which leads to confusing
situation depending on whether the customer turns its modem on
before or after its CPE.

You may have a point here.

I would rather like the text to recommend to only use ULA when
nothing else is available and immediately renumber to GUA when those
are acquired.

And break existing connections on the local network? Total no go.

--
Rémi Denis-Courmont
http://www.remlab.net/