Re: problem statement [was Re: New Version Notification for draft-hinden-ipv4flag-00.txt]

[David Farmer <[e-mail]>]

On Tue, Nov 21, 2017 at 8:32 PM, Brian E Carpenter <
[e-mail]> wrote:

> On 22/11/2017 13:42, David Farmer wrote:
> > On Tue, Nov 21, 2017 at 6:15 PM, Brian E Carpenter <
> > [e-mail]> wrote:
> >
> >> On 22/11/2017 11:33, Nick Hilliard wrote:
> >>> Lorenzo Colitti wrote:
> >>>> On Wed, Nov 22, 2017 at 12:47 AM, Nick Hilliard <[e-mail]>
> >> wrote:
> >>>>> There would also be a requirement
> >>>>> for the network edge to be able to filter out RAs with this option,
> >>>>
> >>>> Why is it not sufficient to configure the routers that emit the RAs?
> >>>
> >>> because you're assuming congruity between ipv4 and ipv6 connectivity on
> >>> a network, which is not codified in any ietf document that I'm aware
> of.
> >>>  If there is a incongruity between management of the two protocols,
> then
> >>> the l2 network must have a mechanism to stop an ipv6 gateway from
> >>> attempting shutting down ipv4 services.
> >>
> >> This is a very interesting point. If every IPv6 router on a link
> believes
> >> there is no IPv4 router/DHCP service, they would all send flag==1 in the
> >> proposed solution. So the proposal only works if at least one IPv6
> router
> >> is colocated with an IPv4 router, and therefore knows to send flag==0.
> >> Oops.
> >>
> >
> > This is a problem only if the flag is set automatically and their is no
> > mechanism to override the automatic setting.  Most routers with an IPv6
> > address configured on an interface by default will send an RA unless you
> > configure them not to.  So, an IPv6 router without IPv4 configured on the
> > interface should set this flag, unless configured otherwise.  In other
> > words you need to be able to tell an IPv6 router there is IPv4 on the
> > interface even if router is not providing it.
> >
> > It is quite possible that the IPv6 topology is incongruous from the IPv4
> > topology, however it is doubtful than the management is also incongruous.
>
> Yes, but do you think we should push a mechanism that assumes correct
> configuration by humans in order to work reliably?
> (Arguing against my own proposal, not for the first time ;-).
>

If you have the funds to build two separate infrastructures, which is what
we are talking about here, one for IPv4 and separate one for IPv6, you
should have the funds to pay for humans smart enough to manage the two
separate infrastructures.  An IPv6 topology incongruous from the IPv4
topology is not going to be a common deployment model, at least for anyone
who will be relying on auto-configuration.

-- 
===============================================
David Farmer               Email:[e-mail]
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota
2218 University Ave SE        Phone: [phone]
Minneapolis, MN 55414-3029   Cell: 612-812-9952
===============================================