On Fri, Dec 16, 2005 at 09:12:26AM +1100, Alexander Samad wrote: > On Fri, Dec 16, 2005 at 01:00:32AM +1100, Anand Kumria wrote: > > On Tue, Dec 13, 2005 at 03:37:09PM +1100, Alexander Samad wrote: > > > > > > I also have 2 internet connections (1 static and 1 dynamic), I onyl use > > > the ip6to4 addressing > > > schema, which is a pain for the dhcp internet address cause I have to > > > make changes to my internet dns, if memory serves me correctly you get a > > > > If you have a static IPv4 address you can use either the AARNet IPv6 > > tunnel broker (reasonably okay) or 6to4 address (tends to have higher > > latency due to dog-leg routing). > > > > If you have a dynamic IPv4 you really want to be using the AARNet IPv6 > > broker. > > I have a bunch of script run to update dyndns, ipsec and a few other > things, so just tacked on a bit to update the ipv6 address associated > with the dynamic
Well, that means that your IPv6 prefix is being updated each time you get a new address. There is lots of pain associated with that (renumbering is easier with IPv6 but on-the-fly router renumbering is still being worked out). > > The site-local prefix (fe80) has been deprecated (rfc3879), instead you > > want IPv6 local addresses (rfc4193) which you can self-generate with > > tools such as: > > http://www.hznet.de/tools/generate-uniq-local-ipv6-unicast-addr.sh > So much reading to do and so little time, just to save me time reading > the rfc's can I still use those address, i will get around to it, but > haven't had a need to follow up in ipv6, but maybe now I do 8) Basically: RFC3330 (or RFC1918) specifies a number of 'private use' IPv4 addresses: 10.0.0.0/8, 172.16.0.0/16 and 192.168.0.0/16 Strangely enough most people and organisations that deploy private-use IPv4 address tend to pick the same ones (so 192.168.0.0/24 and 192.168.1.0/24 are very common). If you then have to connect two 192.168.1.0/24 networks together you end up having to do double-NAT and installing lots of 'funky' routes in various routers. There are a lot of other issues (described in RFC3879) as well, so the solution is to allow people to generate their own prefix but which isn't routed. So long as the generated prefix is unique, you can then connect two (or more) 'private use' IPv6 networks together without having to worry about address collision. The shell script just generates an address prefix for you to use. > > > I routing my encapsulated traffic via ::192.231.212.5 (aus 4to6 gateway) > > > and via the ::192.88.99.1 which is defined as a 4to6 gateway which is > > > mean to be provided by your isp > > > > You mean 6to4, however the 6to4 anycast address (::192.88.99.1) is > > normally provided by the topologically closest network -- at the moment > > that is switzerland as neither AARNet nor Telstra advertise reachability > > of their 6to4 service. > yep, I found one through japan as well, I use it as a fall back Think about it. The one in Japan (I'm assuming you are talking about kddilab.6to4.jp) _is_ advertising itself via anycast. The fact that you aren't using it when you use 192.88.99.1 means that it is topologically closer to route via switzerland than japan. Picking that one means you are increasing your own latency. > Sounds like you have been playing with it for while ? Yes > for work or fun Both Regards, Anand -- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
