[PLUG] IPv6

[Ian Burrell]

Darkhorse wrote:
> 
> Yes of course.  Thing that's curious though is 
> that doing IPv6 over IPv4 apparently requires 
> using more bandwidth to send out the longer 
> addresses.  Isn't 32 million possibilities 
> enough?  Why does the whole world need to 
> be fit into one IP space anyways?
> 

IPv4 is 32 bit, so 4 billion potential addresses.  Except that all of 
those addresses won't be used at once; the usage is fairly sparse.  One 
advantage of a huge address space is that anyone can talk to anybody 
without having to worry about NAT.  Also, it makes the trouble of 
getting addresses from an ISP disappear.

> Should DNS servers be upgraded to report ipv4
> addresses in ipv6 format, or do most of them
> convert on the fly already?
> 

There isn't much point in reporting IPv4 addresses as IPv6 format.  IPv6 
only hosts are pretty rare.

> I wonder if the physical nic hardware has to
> have a big enough buffer to send out ipv6
> packets.  I guess if the MAC addresses haven't
> changed, maybe not.  I'd imagine that those
> have been lengthened too though.  I imagine 
> a 128 bit address space requires a very large
> routing table.
> 

IPv6 headers are 4 bytes smaller than IPv4 headers.  There is overhead 
with encapsulating IPv6 packets across an IPv4 network.

The Ethernet MAC addresses are 48-bit and there are no plans to change 
them.  48-bit is big enough for a while.

One of the big advantage of IPv6 is that could potentially reduce the 
size of the Internet routing table.  Prefixes could be assigned based on 
routing instead of whatever blocks are available or were assigned 
historically.

  - Ian

-- 
ian at znark.com
http://www.znark.com/