Back to index
IP Next Generation Overview
Robert M. Hinden
One-line summary:
As the IP(v4) address space is nearing full allocation, routing and address
assignment are becoming challenging. The paper outlines important features
of the next generation protocol, IPng (or IPv6), utmost of which is the
ability to smoothly transition from the current system to the new.
Overview/Main Points
- We're running out of room in the current (IPv4) address space. That's
not to say that there are 232 hosts out there, but there were
tradeoffs made to make routing easier (class A/B/C/... addressing) that cause
many addresses to be unused.
- As we're squeezing the remainder of the IPv4 address space, routing
is becoming more complicated, and addresses are harder to find. We need
more space, and we need it in the next few years.
- The most important issue in the design of a replacement for IPv4
is that it be interoperable with current systems.
- No ``flag days''.
- Current systems will need to talk to IPv6-enabled systems for
years after IPv6 is introduced, if not longer.
- Some systems, such as printers, that only need to talk to a few machines
may never need to change to IPv6 at all.
- Major changes from IPv4
- Routing and addressing
- 128-bit addresses instead of 32-bit
- Unicast, anycast, and multicast (multicast subsumes broadcast). Unicast
is basically what we know now. Multicast is like today's multicast, but
it has a ``scope'' field to improve scalability. Anycast is like ``send me to
the nearest MCI backbone router''.
- Support for local-use addresses, both link-local and site-local.
Link-local addresses (IP addresses not globally unique, but only unique to,
say, a single ethernet segment) make RARP-like protocols easier. Site-local
addresses (IP addresses unique within an organization) facilitate connection
to the Internet without renumbering.
- Special address formats to indicate IPv4 addresses for compatibility.
- Addresses have hierarchical structure for easy routing. There's lots
of room for that in 128 bits.
- Headers and Options
- Remove or make optional little-used headers. The IPv6 header is only
twice as large as the IPv4 header, though the address size is four times
as large.
- Options are no longer squeezed into the main header. Options are placed
in separate extension headers. An advantage is that by placing end-to-end
options after routing options, the intermediate hops never have to deal
with them.
- Quality of Service and Flows
- Packets can have a 8 priority levels for non-real-time traffic,
and 8 for real-time traffic. These priorities can help the router decide
what packets to drop in case of congestion.
- Packets can be grouped into ``flows''. Each packet in a flow must have the
same values for certain parts of the header, including source and destination
addresses, prority, and some options. Becuase of this, routers can cache
routing information based on flows, instead of having to parse the information
in each packet.
- Authentication and Privacy
- Facility has been made for encryption, signing, and authenticating data.
Algorithms of different strengths and speeds can be used for different
purposes. All conforming IPv6 implementations MUST support DES for
encryption. Put that in the US government's pipe and smoke it.
Relevance
Despite the attempt to make the transition easy, the word from the backbone
network operators is that they're unlikely to switch over any time soon.
The new routers will likely be expensive, and there's no incentive for them
to do it.
Flaws
- Very high level (but that's to be expected from CACM).
- The paper does not address how large the trials that have been done were.
As the scale of the Internet is the main driving force for IPv6, it is
vital that implementations (of which there are a few) scale well.
Links
- IP Next Generation Overview
- http://playground.Sun.COM/pub/ipng/html/INET-IPng-Paper.html
- IP Next Generation (IPng)
- http://playground.Sun.COM/pub/ipng/html/
Back to index