Re: Why has RFC 4941 been designed in such a way, that it might causeaddress conflicts?

[Brian E Carpenter <[e-mail]>]

On 2011-03-21 11:00, Hemant Singh (shemant) wrote:
> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of
> Brian E Carpenter
> Sent: Tuesday, March 15, 2011 6:55 PM
> To: james woodyatt
> Cc: [e-mail]
> Subject: Re: Why has RFC 4941 been designed in such a way, that it might
> causeaddress conflicts?
> 
> 
>> If you'll excuse an anecdote, while I was living in Geneva I was
> regularly
>> amused when the shiny new information screens in the shiny new buses
>> would display the Windows IPv4 duplicate address warning instead of
>> the next bus stop. Even so, the Geneva bus service hasn't come to an
> end.
> 
>> My point? The probability of a duplicate address in an IPv6 subnet is
>> many orders of magnitude less than it is in IPv4. Like 1 in 2^63
>> instead of 1 in 2^8.
> 
>> This is just too remote a probability to worry about.
> 
> In a recent IPv6 CE Router Interop in the U.S. during mid-February 2011
> where such home routers were tested in a cable broadband network IPv6
> link-local addresses were found to be duplicate.  My guess is that the
> Interop network had 2-30 IPv6 nodes.  Two different CE routers with
> different mac-addresses created the same IPv6 link-local address.  I
> don't have logs from the problem but the problem can only be one of two
> things.  Either the CE did not generate its IPv6 link-local address
> using EUI-64 or the CE did but the CE has a bug in the code to generate
> an IPv6 link-local address using the EUI-64 format that uses the
> mac-address of the CE.  The CE routers are consumer devices that will
> deploy with no console.  So what admin of the network will see the CE
> reporting to the CE console a DAD failure for the IPv6 link-local
> address?   The SP serving this home is also clueless as to what
> happened!  
> 
> Stuff happens.  

Yes, and paradoxically this illustrates my point. Passing a law that
addresses shall not conflict, or shall only conflict with a probability
of 1 in 2^63, is beside the point.

Others on this thread have reported they want to secure
> their IPv6 networks.  We should collect a list of all issues operators
> of networks and other folks are reporting and see if the existing
> protocols are not able to help, let's see what else we can do.

Good idea, where's the wiki?

    Brian