just a quick comment on DAD-storms below.
Mika Liljeberg wrote:
On Wed, 2003-03-26 at 13:42, Jun-ichiro itojun Hagino wrote:
note the mechanism is called DAD not DIID. this has been discussed to death numerous times, please refer to the archive.
I know. Means people still disagree about it.
how can you prevent random other node to assign P::X (which you are using and verified fe80::X by DAD) on the interface?
If X is a new ID for the other node, it would have to do DAD with fe80::X. If the node alrady owns X, it could skip the DAD.
In any case, I defend my id X against all possible prefix::X, so the other node will no be able to take it no matter what prefix it uses.
DIID does not work in practice.
It does work if everyone follows the same rules.
Besides, if you do DAD for every address, a new prefix appearing in a RA will cause a DAD storm as every node on the link will attempt to configure a new address at the same time.
This sounds just like the many-nodes-moving issue in MIPv6 which we're working on in mobile-ip.
Current 2462/1 policy causes a delay of 0-1000ms for each DAD packet. So while there may be a storm, it shouldn't affect L2.
Additionally, EUI-64 based addresses are only one of a set of available addresses now, with rfc3041 and CGA based addresses being possible. In these cases (manditorily in CGA) different suffixes are required for each address (local, global scope), so there is no gain from doing DIID. With 3041 addresses, we're still likely to see an additional EUI-64 based address for local comms anyway.
I'd make the guess that at least 3041 addresses and possibly CGA-like addresses will be prevalent as IPv6 adoption continues.
Greg
-------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
