[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: draft-narten-ipv6-3177bis-48boundary-02.txt



On 14-jul-2006, at 8:50, Mark Smith wrote:

If everyone accepts that ipv6 addresses have variable length subnet masks
(and everyone has the appropriate tools to handle this)

Actually the SUBNET mask is now pretty much fixed at 64 bits. From RFC 3513:

   For all unicast addresses, except those that start with binary value
   000, Interface IDs are required to be 64 bits long and to be
   constructed in Modified EUI-64 format.

I don't think this was the right decision although in practice using someone else isn't a good idea as you're going to run into trouble with stateless autoconfig, especially the wway things are now, but even in some cases if this is changed to accept variable subnet lengths.

But we were not talking about subnet sizes, but rather end-user assignment size.

Go to fixed boundary allocations for end-sites, and _nobody_ needs
appropriate tools or processes to validate the sizes of requests from
end-sites. The bits we sacrifice in the IPv6 address space to do this
would be a cheap cost compared to the ongoing and continual costs of
managing variable length assignments to end-sites.

Right. Having only a very small number of different assignment sizes makes life much easier for ISPs, because that way, the blocks are interchangeable. If every user gets a tailored prefix size, then ISPs will have more trouble recycling address space that comes back when uses leave.

Also, regular users don't have any idea how much they need so it's better to use a standard size that fits most users and only think about a non-standard size for non-standard users.

On the other hand, I do see the point that hardcoding prefix sizes especially in standards but also in procedures is likely to come back to bite us in the back at some point.

If nothing else,
hopefully the RIR annual fees would go down, because they'd have less
work to do :-)

++ !

I've understood the idea behind the /48 allocation was that it was
enough for nearly everybody - only the largest organisations or
entities would need shorter prefixes. This seemed to me to be
simplicity winning out.

I think introducing a second catagory of allocations, namely /56s, for
most smaller networks / homes etc, would address that "excessive"
criticism, yet still meet well enough a simplicity goal.

I agree, except that /56 is still too large for small / home networks, and I fear that ISPs will want to skimp even more and give out /64s. And as I said before, individual /64s aren't all that useful in practice.

However, I think going to variable sized allocations for end-sites is
throwing the baby out with the bathwater. It increases addressing
administration costs, is much harder to understand for most network
administrators, and is much more error prone.

Agree. There is also the issue that DNS reverse mapping delegations are easier on 4-bit boundaries, although this isn't a big enough issue to be show stopper on its own.