Gert Doering a écrit :
Hi, On Mon, Feb 18, 2008 at 03:38:04PM +0100, Rémi Després wrote:Note also that ISPs typically have their IPv4 prefixes coming one by one, with different lengths. Free's IPv4 prefixes, for example, are of lengths /16, /15, /14, two /11s, /10.he simple solution of your example couldn't apply.It would work just fine. The tables would be a bit less trivial, but to map 6 prefixes of maximum length /10 into 24 bits worth of IPv6 prefixes is well possible.The code to deal with this, instead of being trivial, and with fixed length parameters, would bacome significantly more complex.The code would need to consult a mapping table, indeed. Which is not very advanced magic.
I can only agree, of course. A table IS indeed possible. It can replace of a simple parameter.Its maximum length will be somewhat arbitrary, but I am sure it can also be lived with. It would make operations and maintenance more complex (address readability in particular), but again, you re right, that's clearly possible.
But the "Keep It Simple, Stupid" principle was a guide. In this instance, I still believe that keeping a KISS design is preferable.Note also that, if 256 ISPs deploy 6rd as is, and if all of them get /24s for that (so that 256 IPv6 subnets are available per site having an IPv4 address). Then, the address space consumed is less than that used for 6to4 (256 /24s, partially used, vs one /16). If this does facilitate IPv6 rapid deployment, in particular because it follows the KISS rule, this seems to me a reasonable price to be paid.
Can we discuss this issue in Philadelphia? RD