Re: [vserver] assigning less than /64 to individual guests
Michael Richardson <[email protected]>
| Newsgroups | gmane.linux.debian.devel.ipv6 |
|---|---|
| Message-ID | <[email protected]> |
>>>>> "Eugen" == Eugen Leitl <[email protected]> writes: Eugen> Just sent that to the vserver list, but figured this is at Eugen> least as relevant. Eugen> So am I in the clear to parcel out a /64 in /80s, as long as Eugen> all the /80 are all on the same LAN or VLAN? No autoconfig Eugen> breakage ensues? Is 48 bits really enough for anybody? If you are saying that you will allocate /80s to each vserver, but actually they will not be layer-2 isolated from each other, this is just an administrative partition, then I think it's a good idea. You can't use autoconfig for the parts that are in the /80s. You can arrange for your /80s to never overlap the autoconfigured stuff. OUI-64s, when generated from OUI-48s (i.e. mac addresses) always have ff:fe as the middle 16 bits. In addition, bit 6 in the OUI-64 (which is bit 1 in little bit-endian, the bit after the "broadcast" bit) will be set if the OUI-64 is believed to be globally unique. Note that if you have IPv6 Privacy Extensions on, then autoconfig will not necessarily set bit 6. I would have to lookup to determine in fact privacy extensions will set the middle 16 bits in any predictable fashion. Permitting autoconfig to work seems like a nice thing to retain. -- ] He who is tired of Weird Al is tired of life! | firewalls [ ] Michael Richardson, Sandelman Software Works, Ottawa, ON |net architect[ ] [email protected] http://www.sandelman.ottawa.on.ca/ |device driver[ Kyoto Plus: watch the video <http://www.youtube.com/watch?v=kzx1ycLXQSE> then sign the petition.