Declaring IPv4 Historic

I've posted a new internet-draft, "IPv4 Declared Historic". I thought I would describe the document, and why I chose to say what I said.

First, the definition of historic:

A specification that has been superseded by a more recent specification or is for any other reason considered to be obsolete is assigned to the "Historic" level.

This doesn't necessarily say that the protocol has been superseded, but the specification. IPv6 is obviously a newer version of the Internet Protocol specification than IPv4. In the sense of "taking the place of a thing perviously in authority," or "supplanting," the IPv6 specification is designed to replace the IPv4 specification. This is precisely conveyed in the following sentence, "IP version 6 (IPv6) is a new version of the Internet Protocol, designed as the successor to IP version 4 (IPv4) [RFC791]."

It's worth noting that I refer to the internet-draft promoting IPv6 to Full Standard, RFC2460bis. This is a dependency; a Full Standard cannot be superseded by a lower level document.

"IPv4 is therefore Historic." As a specification, by the IETF definition of "Historic," it is historic. The implications of "Historic" are that the IPv4 specification (RFC791) cannot be updated. This is actually not as big a step as it sounds like, since the last document to update RFC791 was RFC6864, which in 2013 updated the IPv4 ID field (source: RFC Editor).

In the draft, I say, "IPv4 has inherent limitations which can not be mitigated." The clearest of these is the size of the address space, but others may argue the improvements offered by extension headers, flow labels, and related benefits such as neighbor discovery. "Current and future work builds on IPv6," as described in more detail below. IPv6 is "better for every purpose than the old protocol." There is no application for which IPv4 is more appropriate, and many for which it is not as good, based on address exhaustion and NAT.

The part that I think is most debatable is "The use of IPv4 is deprecated." In order to make this statement, I had to define "deprecated:"

(deprecated:) a feature, characteristic, or practice that should be avoided, in this case because it is being superseded by a newer protocol.

This is based on several definitions, but especially Wikipedia. I further qualify this definition by saying:

The term does not indicate that the practice is harmful, but that there will be no further development in IPv4, and therefore those using the old version are advised to transition to the newer version.

I know some people support this statement, some want an even stronger statement, and others think it is too strong. More discussion is needed to understand where IETF consensus is (if this is to be published by the IETF, it must reflect IETF consensus).

I immediately mitigate the statement deprecating IPv4 by saying that moving the standard to Historic does not mean it cannot be used. Deprecating it means, at most, that we advise against it. I do note that other specifications relying on RFC791 will be affected, but not obsoleted, I think.

To further support the intent of this document, which is to say that the IETF will no update IPv4, I add that "The IETF will no longer work on IPv4 technologies, including transition technologies." This is not an inherent outcome of declaring IPv4 Historic, but an additional requirement. It is my opinion that we have enough transition technologies, and that any other work to improve IPv4 is effort wasted on a protocol that should be on the decline. I also think that in most cases, by the time any new specification is rolled out, IPv4 utilization will be so low that it will be irrelevant.

I note that IPv4 bugs may be discovered, although it seems unlikely that an undiscovered fundamental deficiency in the Internet Protocol of 35 years has yet to emerge. Still, in principle, one should move away from protocols that cannot be updated.

As a final note, I say that IANA can continue its work in IPv4. I note that this is not going to be an issue for very long: in September, IANA will hand out a /18 equivalent to each RIR, with no single block larger than a /20. Allocations beyond that are imporant to the organizations to receive them, but not significant to the large scale Internet.

In summary, there are a few points in this draft that I know need further discussion. The primary statement, though, that IPv4 is Historic, is true according to the IETF definition.

Although I always welcome comments on Twitter, the right place to debate this draft is on the sunset4 mailing list.

Lee Howard
14 March 2016