Speaking of Standards

Carrying on a fine tradition, the IETF today released RFC 6214, which adapts the venerable RFC 1149 (from 1990) for IPv6.

If you are not aware, RFC 1149 defines, of course, IP-over-carrier-pigeon.

RFC 6214 does outline some of the challenges with IPv6 with regard to avian carriers:

As noted in RFC 1149, the MTU is variable, and generally increases with increased carrier age. Since the minimum link MTU allowed by RFC 2460 is 1280 octets, this means that older carriers MUST be used for IPv6. RFC 1149 does not provide exact conversion factors between age and milligrams, or between milligrams and octets. These conversion factors are implementation dependent, but as an illustrative example, we assume that the 256 milligram MTU suggested in RFC 1149 corresponds to an MTU of 576 octets. In that case, the typical MTU for the present specification will be at least 256*1280/576, which is approximately 569 milligrams. Again as an illustrative example, this is likely to require a carrier age of at least 365 days.

Furthermore, the MTU issues are non-linear with carrier age. That is, a young carrier can only carry small payloads, an adult carrier can carry jumbograms [RFC2675], and an elderly carrier can again carry only smaller payloads. There is also an effect on transit time depending on carrier age, affecting bandwidth-delay product and hence the performance of TCP.

I did very much enjoy this part of section 5:

Routing carriers through the territory of similar carriers, without peering agreements, will sometimes cause abrupt route changes, looping packets, and out-of-order delivery. Similarly, routing carriers through the territory of predatory carriers may potentially cause severe packet loss. It is strongly recommended that these factors be considered in the routing algorithm used to create carrier routing tables. Implementers should consider policy-based routing to ensure reliable packet delivery by routing around areas where territorial and predatory carriers are prevalent.

There is evidence that some carriers have a propensity to eat other carriers and then carry the eaten payloads. Perhaps this provides a new way to tunnel an IPv4 packet in an IPv6 payload, or vice versa. However, the decapsulation mechanism is unclear at the time of this writing.

All in all a fun bit of work… enjoy!

Related posts:

1. The challenge before Speermint and Drinks – moving beyond full mesh into open interconnection

Tags: IETF

You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.