Summary

  • Haisheng Yu argues that an IPv6 plan should be legible to the people who operate it: region, service and function can occupy clean hexadecimal boundaries instead of being packed across them.
  • The APNIC proposal he wrote with Haijun Li would let an eligible LIR request nibble-aligned allocation after documenting its need. APNIC 62 reached meeting consensus; the proposal still has a policy-process path ahead.

An IPv6 address can contain more room than a network engineer will ever want to count. A useful address plan can still run out of something much rarer: patience. Haisheng Yu's APNIC essay makes the case for spending a little of that abundant address space to save human effort. He wants the structure visible in the hexadecimal address itself, where each character represents four bits, or one nibble.

The proposal has a person behind the geometry. APNIC identified Yu as a CNNIC participant asking a question at an APRICOT 2026 IPv6 session. In July he explained his planning method in his own name. Put a region at one clean boundary, then divide its contiguous block by service: an operator can recognise what it is looking at and preserve a route that covers the region. Reverse the order, scattering one city's networks across service pools, and the neat city route becomes several announcements. His /32-to-/40-to-/48-to-/56 example is an illustration, not a demand that everyone copy the same blueprint.

Yu also describes a region field squeezed into five bits. It has room for 32 values; the thirty-third region sends the designer back to the drawing board, potentially with renumbering attached. A whole byte would have bought headroom. The joke in a huge address space is that the cheapest thing to save can become the most expensive thing to manage.

With Haijun Li, Yu carried this thinking into APNIC prop-170's original text. The proposed Section 8 mechanism concerns allocations to local Internet registries. An applicant would explicitly request nibble alignment and present a long-term address plan; APNIC would assess the underlying requirement first, then normally allocate the smallest aligned block that meets it. The /32 minimum and ordinary one-bit subsequent doubling remain. Existing holders face no blanket renumbering. If contiguous expansion fails, the text offers a separately justified allocation rather than pretending adjacent space exists.

APNIC's assessment says its tools and guidance would need changes. The APNIC 62 report records consensus at the meeting and further policy steps. Yu's interesting move is narrower than a victory lap: he gave a human-readable network design a route into a documented allocation request. The next test is whether that route asks for real plans, rather than treating four-bit tidiness as a free ticket to a larger block.