To mark APNIC’s 30-year anniversary, the APNIC Blog is running a series sharing stories, anecdotes, milestones and insights that capture some of the essence of the last three decades.
We’d love to hear your stories and memories of the APNIC community. Post them on social media using the tag #APNIC30th!
At the birth of the registry model, Internet engineers could already foresee that the supply of IPv4 addresses would not sustain the predicted rate of consumption. At the time, address allocation was based on three fixed sizes: Class ‘A’ addresses, which had 1.6M hosts, Class ‘B’, which had 65,000 hosts, and Class ‘C’, which had 255.
The first Regional Internet Registry (RIR), RIPE NCC, was launched in 1992 and during its first 10 years, the equivalent of roughly 15% of the usable IPv4 address space had been delegated to around 6,800 participants worldwide.
When APNIC launched in 1993 this increased to nearly 18% of the usable address space delegated to over 12,000 entities within a year. The rate of participation had almost doubled, and the amount of address space being consumed had increased by 20% in a single year. It became apparent quickly that maintaining this rate of growth with class-based addressing wasn’t possible.
‘Classful‘ addressing based on these three categories (there were five classes, but two of them aren’t relevant here) of address size was far too coarse. There were simply too many eligible entities seeking to come into the Internet to sustain this ’three size‘ model.
Andy Linton recalled the nuances of classful addressing.
“I applied for my first IP address block in 1987 and it was documented in RFC 1020 …search in the document for my NIC handle of AL46. The process was straightforward — you sent an email showing that you worked at a university, and you were assigned what was then called a Class B address (/16). I was working at the University of Newcastle upon Tyne, and we were assigned 184.108.40.206/16. Had we applied 20 years later we wouldn’t have been allocated anything like as much space!” Andy said.
Classless Inter-Domain Routing (1993)
By September 1993, the basic model of how to convert from the classful A/B/C model to a more nuanced classless basis of address distribution was understood. It was documented in RFC 1519:
We believe that the judicious use of variable-length subnetting techniques should help defer […] the exhaustion of the 32-bit address space. […] In should also be noted that the present method of flat address allocations imposes a large bureaucratic cost on the central address allocation authority. For scaling reasons unrelated to address space exhaustion or routing table overflow, this should be changed. Using the mechanism proposed […] will have the fortunate side effect of distributing the address allocation procedure, greatly reducing the load on the central authority.
Classless allocation solved one part of the problem, but as with most solutions, it created a whole new set of complexities and a need for innovative, appropriate processes.
This change represented the beginning of address policy, but it also asked some significant questions, namely, how can an applicant justify their request for addresses, and how to decide how much to delegate?
Adoption of the rough consensus process for address policy
The emergent behaviour was to adopt the IETF ‘rough consensus’ model. Rough consensus sought to explore the community’s range of views and adjust policy accordingly, noting divergence from a core body of belief. This model explicitly drove compromises that reflected the majority view, with a preference for strongly evidenced decisions.
Underpinning this approach was a widely held philosophy. Andy Linton, for instance, recalls these ideals were fundamental to establishing open membership organizations like InternetNZ as “a totally independent entity, which operates within the broad structure of a not-for-profit service operation and applies a single community policy in an open and fair manner”.
“I’m not sure who came up with this wording but I think it caught the spirit of the times and has continued to be a model for how this part of Internet governance has operated,” Andy said.
Even today, APNIC’s current consensus process reflects the experience gained since 1993 in applying this approach to community-led decision-making.
Consensus processes are, by nature, convoluted. They require active engagement through mailing lists and informal consensus-seeking both in written communication and face-to-face interactions. Then, they move on to a more structured consensus-building process during in-person meetings, followed by the formalities of review and discussion. Significant objections can be restated, potentially leading to the reversal of consensus decisions. Consensus, more often than not, emerges as a somewhat imperfect outcome, especially in the context of challenging decisions.
“In the very beginning of APNIC, there was a mailing list of folks I tagged to help in such decisions. I tried to get at least one person from each (active) economy to participate. If there was a new scenario, I’d float it on the list and ask for opinions. Getting consensus wasn’t really a target, rather it was getting someone (anyone) to voice their opinion,” remembered David Conrad, the first Director General of APNIC.
The process has since matured and will continue to evolve.
“Eventually, I wrote the resulting opinions into APNIC-001 and then most of the policy decisions were codified in what eventually became RFC 2050 (although by the time that RFC was published, it was already out of date). It took quite a while before folks began to express their opinions about address policy sufficiently for the concept of ‘(rough) consensus’ would apply,” continued David.
During the early stages, it became evident that consensus building in the Asia Pacific region required different solutions to other regions.
“To be honest, the idea of ‘rough consensus’ didn’t really translate well into the Asia Pacific region,” said Anne Lord, who joined APNIC in 1998.
“Hierarchy and status play such an important role. I’d experienced the RIPE community and it was obvious there were clear differences to the APNIC community, who were just very… quiet. So, it was often the ‘Westerners’ who spoke loudest; decisions at meetings therefore only reflected a subset of opinion. The rest were lost in translation or too afraid to speak or put their hand up.”
According to Aftab Siddiqui, cultural norms play a significant role in governance.
“The Western expectations of process are rooted in their established governance structures and decision-making mechanisms, while some Asia Pacific economies may align closely with these practices, others have unique governance models that suit their cultural, historical and/or political contexts, leading to different decision-making processes,” said Aftab.
Akinori Maemura suggested there was learning required by all parties involved.
“I actually struggled to fit myself into the rough-consensus decision-making process. It was completely different from how we had been dealing with decisions in our own culture — talking with each other to ‘check temperature’ and seek consensus before the public discussion began. We change our style for IP policies — to propose it publicly, voice our idea publicly, give feedback publicly,” agreed Maemura.
“I do remember JPNIC asking a million and one questions regarding the nuances of how elements of the process were to be interpreted. They managed to come up with all kinds of clarifying questions relating to scenarios we had not really considered,” Anne said.
Towards a culturally sensitive ‘rough consensus’
Given that this early version of ‘rough consensus’ wasn’t innate for all cultures across the Asia Pacific region, how did APNIC seek to achieve a more inclusive approach? The short answer is trial, error, and adaptation.
Partly, it was building social bonds within the community.
“One of the reasons Barry Greene and I started up APRICOT was to try to replicate the informal/personal networking/social aspects of those meetings as a way of doing a tech transfer into the AP region that would be ‘more friendly’/less confrontational than you’d typically see in those meetings in the US and Europe,” recalled David Conrad.
Removing ambiguity and clear expectations also helped.
“The lesson learnt in the APNIC region was that there was a strong need for a very clear set of rules regarding the policy process, more so than in other regions. There was no room really to be wishy-washy — once the process was in place. This made some people more comfortable about participating, I think. It has taken a lot of work over the years by APNIC to support the community to find its voice,” Anne observed.
Encouraging discussions that didn’t require standing in front of a crowd also proved helpful.
“Using the mailing lists in addition to the meetings really gave people the time and space to make comments since some sections of the community were often more comfortable with supplying written comments,” Anne said
While translation and remote participation help people feel comfortable speaking up, much of the work happened developing new norms and habits in local and national groups.
“As a part of the founding JANOG Committee, we tried our best to get people to line up to the floor microphone rather than keeping quiet at their seats with smiles. It is indeed successful, and you see many people in the queue up to the floor mic,” Maemura said.
“I am really happy to see many people put their proposal forward and the multitude of proposers to Policy SIG these days,” he added.
30 years of address policy
Policymaking is a gradual, ongoing, and imperfect process that continues to be adapted to the needs of the community.
“The Internet landscape in the Asia Pacific region has seen significant growth and transformation over the years. Multiple economies in the region have increasingly engaged in policy discussions and other related initiatives, they have been able to express their unique perspectives, concerns, and priorities,” Aftab reflected.
While there’s still work needed to thoroughly represent all the voices in the region, due to these efforts, participation is now much broader and more representative than when it started.
Address policy has grown outwards from the purely technical and increasingly reflects social considerations and legal requirements that weren’t considered early in the life of this policy process. The strength of the APNIC (and wider RIR) address policy lies in understanding diverse perspectives.
“… finding a cohesive and representative voice requires time, engagement, and collaboration among stakeholders,” Aftab resolved.
In many ways, the criticisms of any consensus-seeking approach become its strength!
The quotes in this article were sourced from private communications or public postings. Some have had minor edits for clarity.
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.