Summary

  • Ricardo Patara’s LAC-2012-10 put a four-year ceiling on the addressing plan supporting an initial IPv6 allocation larger than /32 and required each planned POP block to cover at least 30% of that POP’s current need.
  • LACNIC changed “binary boundaries” to “hexadecimal boundaries” at consensus, implemented the policy in March 2013, then removed the four-year and 30% tests through a separate policy implemented in August 2017.

IPv6 has enough addresses to make a spreadsheet feel humble. Ricardo Patara still wanted the spreadsheet. In his 2012 LACNIC proposal, an operator asking for an initial block larger than a /32 had to document why. The request could look ahead, but its addressing plan could run no more than four years. Patara’s rationale was practical: even large companies generally plan investments over three to five years. A request for vast address space therefore had to meet a calendar an engineer and a registry analyst could both read.

The second ruler sat at each point of presence. A network with multiple POPs could set a minimum prefix for each, but the block foreseen for any one POP had to cover at least 30% of that site’s current needs. It was a peculiar fraction in a technology famous for giant powers of two. That peculiarity was its job: make an expansive future topology answer to some present infrastructure. In prefix notation, a block larger than /32 has a shorter numerical prefix; Patara was describing more address space, not a longer slash number.

Then the meeting corrected the label on the ruler. The original English text called /X, where X is divisible by four, a “binary boundary.” Four bits make one hexadecimal digit. LACNIC’s meeting account says consensus came with “fronteras binarias” changed to “fronteras hexadecimales.” An edit of one word made the operating instruction match its mathematics. The policy archive records implementation on 11 March 2013. Patara, then at NIC.br after eight years at LACNIC, had given large initial IPv6 requests a clock, a site check and a legible alignment rule.

The clock did not survive forever. The staff analysis of LAC-2016-7 says its replacement text removed both the four-year limit and the 30% POP condition. Implemented on 16 August 2017, the later rule weighed clients, users, infrastructure, hierarchy, geography, segmentation and expected allocation longevity. This was another author’s proposal, and a different bargain: fewer fixed numbers, more room to explain a network’s shape. Patara’s 2012 test remains a sharp historical question. When a registry asks for “need,” how much should arrive as arithmetic, and how much as a credible map of an operator’s future?