Tunneling Route Reduction Protocol (TRRP)

Implementation: How Do We Get There From Here?

Phase 1

The protocol is standardized.

trrp.arpa is created and delegated to NRO.

Initial ITR software is built and tested. Note that all software needed to support TRRP except for the ITR software is already widely deployed.

RIR policy is proposed and passed for making Micronet end user IPv4 assignments of /25 and longer and IPv6 assignments of /56 and longer.

Phase 2

Carrot: ARIN and the other RIRs begin making Micronet assignments. These micronets will never be directly routeable on the Internet; instead they must use TRRP.

ARIN contracts one ISP to provide a route-of-last-resort for the the entire group of Micronets which leads to an ITR. The ISP agrees to provide 64kbps of tunneled bandwidth to each and every Micronet in exchange for the exclusive right to sell additional bandwidth to those who want it. As ITRs are not yet widely deployed, this is a pretty good deal.

Phase 3

Carrot: ISPs notice that by implementing a local ITR they get better connectivity with servers hosted on the Micronets.

Carrot: ISPs notice a competitive advantage to offering an ETR for their customers so that they can use Micronets without needing their own ETR. This is productized into a service which offers the ETR, the DNS Route server and negotiation with ARIN to get the Micronet assigned.

Carrot: Software vendors notice that the Micronets offer a growing market for a better DNS route server than static Berkeley Bind.

Phase 4

Stick: Fed up with the continued growth of the BGP routing table, one of the Tier-1 backbones announces that between 3 am and 4 am headquarters local time they will filter and discard any non-customer prefixes longer than /23. Folks advertising a /24 are advised to implement an ETR and a DNS Route server in order to retain connectivity. *Note* that a few /24's are used for DNS servers at the root of the TRRP hierarchy. Those /24's can not be filtered.

Stick: A second Tier-1 ups the ante by filtering between 1 am and 7 am. At this point, end users with /24's and their small ISPs wake up, implement the simple ETR and insert a route into their DNS.

Stick: Some sites without ITRs notice that they can't reach a bunch of the Internet during the middle of the night. They implement (or pester their ISPs to implement) an ITR to fix the problem.

Stick: In order to avoid a costly upgrade, a third Tier-1 finally decides that it will filter prefixes longer than /23 period. The remaining DFZ participants quickly follow suit.

Carrot: Open source software authors notice that by implementing a TRRP ITR directly in the network stack on the originating host, their servers have better connectivity to folks with prefixes longer than /23.

Phase 5

A repeat of phase 4 until prefixes longer than /20 are not advertised in the DFZ.

Stick: Noticing that not that many IPv6 networks have been deployed yet and that they've all implemented ITRs, the Tier-1's refuse to route IPv6 /48 PI space, requiring folks with PI space to use TRRP instead.

Carrot: Microsoft notices that Linux has better connectivity to the Internet because it implemented TRRP ITR in the network stack on the originating host.

Phase 6

IANA establishes an AS registry where each AS may specify the prefix it has designated to be Primary Space.

Carrot: Small endpoint AS's implement DNS Route servers for the address space they didn't designate as primary as a preemptive measure: just in case someone decides to further filter.

Stick: If the growth of the BGP table hasn't already stabilized to something reasonable, someone decides to further filter.

Carrot: Seeking a more efficient and effective tunnelling mechanism than GRE, multiple software vendors offer alternate tunneling protocols. One or two emerge as victors and see widespread deployment.