Re: [v6ops] FW: ULA & Basic Requirements for IPv6 Customer Edge Routers

[Mark Smith <[e-mail]>]

On Mon, 13 Dec 2010 14:03:37 -0600
"Hemant Singh (shemant)" <[e-mail]> wrote:

> FYI to v6ops.  Also, James, please note that Brian Carpenter already
> said that the ULA text in the IPv6 CE router document was already agreed
> upon (with rough or whatever consensus) and then the document went past
> the IESG and to the RFC Editor queue.  So we could avoid debating the
> existing ULA text in the document.  The only debate is related to new
> text which relates to the coexistence problem with the solution to the
> problem as the MSR in the RA for the ULA.  As Fred also says, any other
> issue can be taken up in the IPv6 CE router bis document.  
> 
> The reason I am sending the email below to a wider audience is to bring
> out one key fact.  The fact being that the private IPv4 and IPv6 ULA
> coexistence problem causing Internet access delay for the host in the
> home is NOT merely a coexistence problem with the ULA.  The problem also
> exists with the IPv6 GUA (Globally Unique Address).  See below.
> 

I've been wondering a bit why ICMPv6 destination unreachables didn't
solve the problem, but figured I didn't completely understand the
the problem yet.

I don't think announcing a router lifetime of zero solves it for the
ULA only case, as I think it would prevent inter-LAN interface routing
between ULA subnets in a ULA only scenario. I'd think nodes would
consider every off-link destination to be unreachable if they don't
consider the CPE a default router.

I think it would also think the coexistence issue, in the absence
of ICMPv6 destination unreachables, would occur for ULA only
devices when there are some LAN interfaces with a global + ULA prefixes
and others that are ULA only. The current RA lifetime zero text wouldn't
apply in that situation.

Regards,
Mark.


> Thanks,
> 
> Hemant
> 
> -----Original Message-----
> From: Hemant Singh (shemant) 
> Sent: Thursday, December 09, 2010 6:53 PM
> To: james woodyatt; Ole Troan
> Cc: [e-mail]
> Subject: RE: [v6ops] ULA & Basic Requirements for IPv6 Customer Edge
> Routers
> 
> James,
> 
> Not quite.  Please see in line below.
> 
> -----Original Message-----
> From: james woodyatt [mailto:[e-mail]] 
> Sent: Thursday, December 09, 2010 5:54 PM
> To: Ole Troan
> Cc: [e-mail]
> Subject: Re: [v6ops] ULA & Basic Requirements for IPv6 Customer Edge
> Routers
> 
> 
> >Actually, I think my disagreement is much narrower than that.
> 
> >1) I think recommending the advertisement of only a ULA prefix on LAN
> interfaces, when the router is dual-stack and also advertising IPv4
> public Internet service with NAT and RFC 1918 addresses, >directly leads
> to the coexistence problem that I identified in
> <http://www.ietf.org/mail-archive/web/v6ops/current/msg06363.html>, and
> that this IPv4 coexistence problem is too serious for IETF to allow
> >this RFC to be published at this time with any such recommendation.
> 
> This problem was reported to us on Wednesday evening (EST) before
> Thanksgiving and in one week or less we had a solution to the problem.
> The same Wednesday the first response we gave was to have the host
> implement Happy Eyeballs.  Since one cannot be sure when a host would
> implement Happy eyeballs, we then proposed a solution which was to say
> the IPv6 CE router sends a RIO (MSR of RFC 4191) for the ULA prefix and
> this solution was at least satisfactory to some of the parties who
> contacted us - Tore Anderson is one.  Tore, Lorenzo, and Erik Kline were
> amongst the folks who contacted us with the problem and our thanks to
> them. They worked with us for the new text once the solution was
> devised.  We authors of the IPv6 CE router document also raised a bug in
> the tested IPv6 CE router.  The bug was why didn't the router send an
> ICMPv6 Destination Unreachable to the host on seeing an IPv6 packet with
> a IPv6 Globally Unique Address (GUA) for destination and a ULA for
> source address?  Such a packet cannot be sent out the WAN as per zone
> violation rules in RFC 4193.  We already have some ULA zone boundary
> warnings in the IPv6 CE Router document.
> 
> [In this role, the IPv6 CE router is responsible for ensuring that
> traffic using its ULA addressing does not go out the WAN interface, and
> does not originate from the WAN interface.]
> 
> >2) The authors of the draft seem to believe one or more of the
> following propositions, which I reject:
> 
> >  a) the coexistence problem is not serious enough to warrant
> withdrawing the draft to address it;
> 
> We have a solution to the problem.  So what specifically problem do you
> have with the solution?  Also, on the first day of the problem reported
> to us, I said, this is not just a coexistence problem.  When the host is
> provisioned with an IPv4 private address and an IPv6 GUA, and if the
> IPv6 WAN link (IPv4 link is still active) of the IPv6 CE router goes
> down, the host will send sent packets to the Internet and the host would
> still experience a delay with Internet access. This delay with Internet
> access is the problem of coexistence of private IPv4 address and the
> ULA. 
> 
> Also, note my responses to you in the email message URL shown below.  No
> satisfactory IPv6 solution has been given to us for the 3 problems we
> have always listed where the ULA is needed. 
> 
> http://www.ietf.org/mail-archive/web/v6ops/current/msg06315.html
> 
> 
> Hemant 
> _______________________________________________
> v6ops mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/v6ops