[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: 6rd - Rapid Deployment on existing IPv4 infrastructures - a newapproach
On Tue, 12 Feb 2008 23:52:07 +0100, Rémi Després <remi.despres@free.fr>
wrote:
> On my home LAN, it doesn't suck at all. It just works to my complete
> satisfaction.
> Clearly it is desirable to do much better than this /64 first step, but
> one step after another is sometimes a good way to progress.
The IETF has a long history of doing "provisional" "for a short time"
protocols that eventually last much much longer than intended. NATs are
one. And 6to4... it was standardized over 5 years ago, and we are only now
realizing that it does not work as well as intended. When we know that
something is wrong or broken to begin with, we definitely should not
standardize it in the first place, IMHO. That will not prevent Free from
using it anyway.
> In view of your strong reaction though, the following sentence will have
> IMO to be rewritten:
> "This is expected to be satisfactory initially for the vast majority of
> residential sites, where the number of hosts is small."
The number of hosts is totally irrelevant.
I have been on switched IPv6 subnets with hundreds of boxes without any
problem.
The issue is with *cascading* routers. With no bits to do routing within a
site, if bridging is not possible, there is no choice but to use NAT.
> You are free to believe that IPv6 to IPv6 NATs would be the proposed
> next step.
> But not by me, for sure.
So what if I plug a cheap SOHO router behind my CPE?
> You are right, they could have done that.
> But IMO the additional complexity would not have been justified,
Where is the complexity in shifting some bits?
> especially since it was expected to be only provisional.
> It would also have made addresses less readable.
That assumes humans can convert decimal IPv4 address to hexadecimal fast
enough to "read" them.
Also, this only works for /32+32 and /16+32 prefixes, neither of which look
like a good compromise.
I'd happily preserve statelessness, and get saner prefixes size at the
expense of readability. IPv6 addresses are pretty much unreadable anyhow.
--
Rémi Denis-Courmont
http://www.remlab.net