In her APRICOT 2015 keynote, Elise Gerich from ICANN presented a historical review of the Internet numbering system that led us to where we stand today. Elise is uniquely suited to talk to this subject, having a 22 year history of leadership in the Internet, and her early role in the NSFNet function. Elise provided much of the critical infrastructure which led directly from an Academic/Research network, to the fully commercially viable Internet we have now. Elise did this for MERIT, the agency which had management oversight of the emerging network.
Back in the early days, the ABC’s of networking was literally about the class-A, class-B and class-C networks. We had a grossly simplistic view of the world, and fitted everyone in a three-bucket size model. Back in 1987/1988, Elise was a member of the FEPG, the Federal Engineering Planning Group, which informed the FNC or Federal Network Council, with oversight of all US federal agency networks. For much of the early years, the rate of growth of visible routed IPv4 networks was pretty small, and looked linear. It was possible to believe there was no future exhaustion problem on the horizon. SRI-NIC acted as a central clearing house, and required connecting networks to demonstrate a sponsor from the federal agencies. SRI-NIC was a function (NIC for Network Information Center) run by SRI international, a non-profit from Stanford University. The address allocation model was based on a simple application that flagged which addresses were connected to the research community.
Elise was able to find in the donated historical records (still on paper!) at the computer museum, that as far back as 1988 there was considerable interest in allocating addresses to the Japan research community, who were asking for a large allocation of A’s, B’s and C’s. Some of the history is now lost because of the inevitable loss of paperwork but Elise suspects a few critical people may have enacted the first regional allocation model in the Asia region, well before the RIPE NCC established a European registry.
One of the critical problems back in the eighties was the ‘acceptable use’ conditions which went with a connection application. Getting addresses was one thing, but you had to demonstrate ‘acceptable use’ – some need – to be connected to the NSFNet backbone. MERIT, who ran the backbone, was allowed to accept people who self-certified a need; packets flowed on a promise! This was a moment of change in number allocations, from connected/non-connected status (flagged at SRI-NIC) to routed or unrouted (flagged at MERIT).
Initially in the eighties, only three countries had access to the NSFNet backbone: France, Canada and the USA (the UK used another path into the global Internet of that time, via the SATNET project). By the end of the nineties, many more had joined. In 1990, the Internet Architecture Board (IAB) wrote RFC1174, recommending a principle for IP address allocation, which has substantively continued. This RFC explicitly presaged the distribution of the SRI-NIC function to regional levels. This in effect promoted a concept of the FEPG, into the IEPG, taking the federal initiative into an international initiative worldwide.
In 1991, SRI-NIC’s contract ended and the function was passed to Network Solutions, formalizing the IANA function and constructing a registry role. This was also when the RIPE NCC was formed (in 1992) as an instance of a formally constituted regional registry in line with RFC1174.
The qualities of how to qualify as a regional registry codified in RFC1366, and this also led to the documentation of the sub-allocation of the class-C space (255 host instances) as blocks, assigned to world regions, for sub-allocation by regional registries. The class C space was used, because of caution in the minds of IANA, IEPG and FEPG, concerned about release of the class-A and class-B spaces. The idea might not work, and they were concerned at exhaustion of the initial class A and B spaces.
Network aggregation had been conceived, and so a subsequent RFC also defined mechanisms for the adoption of more complex Classless Internet Domain Routing (CIDR). At the time, around 9,000 routes were in the global default-free routing zone. It’s ironic to look back on this time when now, over 500,000 routes are in the global routing table, but the concern was very real at the time and a lot of time and energy was consumed in routing workshops, leading to RFC1519 which formalized the model of address/mask based routing. This was a point in time when address allocation models had to now translate into engineering changes in the routing hardware and software, so a huge amount of Internet Standards work stemmed from this decision, a 1993 launch date of CIDR led to a huge amount of work we now all benefit from.
It was at this time, that JPNIC came into being, with APNIC formally created around this time and receiving two blocks in 1994 under the aegis of the RFC1366 model. ARIN came in 1997, LACNIC in 2002, and AFRINIC in 2005. The fundamentals of the model have not changed, from these roots.
IEPG has morphed from a formal role, to a much more informal meeting of technical minds. Elise showed a small core attendance at a 1990 meeting, many of whom are still active in more senior roles in regional, national and international networking, academic and commercial. Now, IEPG is a morning session at IETF, without RFC status and conducting an open dialogue with free attendance.
Elise’s talk (below) is well worth reviewing for the details not covered here. An engaging and enjoyable walk back in time, to help explain why we’re here now.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.