On 4/17/2010 6:33 PM, William Herrin wrote:
On Fri, Apr 16, 2010 at 9:57 PM, Scott Brim<[email protected]> wrote:
Unfortunately there isn't much new under the sun.
Matthew allegedly wrote on 04/16/2010 15:07 EDT:
I would like to propose the following concept for discussion. The idea
is to either extend IPv4 or create a new protocol that would work with
IPv4 in order to allow a backwards compatible, yet hierarchical
addressing model.
Hi Matthew,
Welcome!
Though intended as a half-joking "so-there" when someone years ago
claimed there was no conceivable way we could have extended IPv4, I'd
welcome your thoughts on: http://bill.herrin.us/network/ipxl.html
I thought about extended IPv4 and had heard that it was proposed. It
would be a good idea but when I started thinking about my proposal I
realized that a larger address space that leaves IPv4 space intact and
gives a hierarchy would be more ideal. A true hierarchical address
structure would be much more efficient.
96-bit identifier based on region, service provider, etc> <32-bit
IPv4 style address>
In addition, organizations with an ASN would have the following
fully-unique address:
<16 or 32-bit ASN> <96-bit ID> <32-bit IPv4 style address>
One of the things we've discovered over the course of our
investigations here is that overloading multiple functions on a single
number is an often regretted decision. How much less trouble would we
be in today if the protocol stack didn't use the source and
destination IP addresses in the process of associating a TCP packet
with its connection?
The AS number already has a purpose. It isn't forwarding packets.
Even if the ASN portion of this concept was left out we would gain the
benefit of backwards compatibility + hierarchy in the address.
Multi-homing would still lead to increased routing tables but IPv6 has
this problem too, as does IPv4. Although it is conceivable that we could
mitigate some of the table size by allowing multi-homing AS's to use a
portion of the 96-bit ID where most of the duplicate address could be
hidden lower in the hierarchy.
Although we don't have to base this structure on geographic location I
still believe having a system more like house addressing is better in
the long run. However, instead of geographic cities, etc. there are
Service Providers, and possibly large companies which would make up the
96-bit ID. It works on a massive scale on Earth and I think it could be
adapted for IP Addressing. It turns routing into a distributive function
where not every route needs to be on backbone routers. Just the major
Provider's.
I know this is also how we want IPv6 to go but in the end IPv6 is a flat
128-bit space which is then being logically folded into a hierarchy.
This can (and probably will) lead to the messed up tables we see today.
Regards,
Bill Herrin
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
