Philip Homburg wrote: > In your letter dated Sat, 28 Mar 2009 11:38:42 +0100 you wrote: >> And import the complete IPv4 routing table into the IPv6 routing table. >> No thank you. > > No, instead of that, every ISP gets a completely new prefix, wastes a lot of > bits in that prefix (you need 32 bits for the IPv4 address), 6rd does NOT require that. Did you read Remi's draft? If an ISP has a IPv4 /21 for their customers they want to IPv6 enable then they need 32-21 = 11 bits. Thus they need a /48 - 11 = /37 for their 6rd prefix. Then every user gets a nice /48. Or what Free does, give a /56 which is 'good enough for endusers' and thus you only need a /56 - 11 = /45. If they ISP has 5 IPv4 /21s they thus need 5x /45 for this trick. That address space can later, when the users get real native v6 (if ever) be re-allocated for other things. Still only 1 /32 (or larger if the ISP was able to justify that) will be visible in the IPv6 routing tables. Instead of 5 6to4 prefixes which should not be there in the first place. > and then adds it to the routing table. And you can't even give customers > the /48 they are supposed to get. > >> Fortunately 2002::/16 orlonger is pretty well filtered. > > You are saying that if an ISP gets a /32 from a RIR it is no problem, but a > 2002::xxyy/32 suddenly is a problem? Yes, as that is what the RFC specifies. As I wrote in my previous email (and thus still a few lines above here) to avoid importing the IPv4 routing tables into IPv6. IPv6 != IPv4. Please actually read the 6to4 RFC and the 6rd draft. Greets, Jeroen
Attachment:
signature.asc
Description: OpenPGP digital signature