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

[Philip Homburg <[e-mail]>]

In your letter dated Tue, 15 Mar 2011 19:04:48 +0100 you wrote:
>On 2011-03-15, at 16:58 , Philip Homburg wrote:
>
>> I think the answer is that is statistically very unlikely that on a =
>single
>> subnet, a 64-bit random number will ever be equal to any address =
>manually
>> configured in DHCP.
>
>I'd say this entirely depends on how the (usually pseudo-) random number =
>has been created. No RFC says it is forbidden to choose a random IPv6 =
>address by simply picking a number between 1-255 and pad all the bits to =
>the left with zeros. It may not be very advisable to do that, but it is =
>not disallowed either. 

Yes, it is possible that some junior programmer doesn't know how to generate
a 64-bit random number. Hopefully people will figure out that some device is
broken in that regard and get the manufacturer to fix it.

But there is an easy fix for you. What if you put something in those
bytes? If you come up with a number that relates to just your
organisation, then it is very unlikely that a broken random number generator
will generate exactly that pattern.

If you just need stable addresses, then you can also put your own random
numbers in DHCP. 

>How would two ethernet cards have the same MAC address, unless it has =
>been manually configured? Doesn't globally unique mean "globally unique" =
>anymore?

No, it never did. That's the problem. From manufacturing defects (a batch
of cards all with the same ethernet address), small mistakes (a manufacturer
reusing a range) to companies that doen't know how it works and just use
some random vendor id, to cards that can just be jumpered to look some
other vendor's cards because that is what the software expects.