The world in which IPv6 was a good design

At the IETF

Last November an engineer attended an IETF meeting and came back with a useful piece of perspective: the IPv6 debate still rages. It has been reported that many attendees were bullish — convinced IPv6 would finally replace IPv4 “Any Day Now” — while others treated it with suspicion. The same meeting introduced TCP BBR, which drew cautious optimism; sometimes the new thing is actually an improvement, and sometimes it’s too-good-to-be-true. Which is IPv6? Depends who you ask.

Layered history explains the mismatch

The short version: IPv6 isn’t just “IPv4 with more bits.” The protocol landscape it tried to address grew out of two very different engineering worlds — long-distance, point-to-point circuits (telephone-style leased lines) and local bus-style LANs — and those roots shaped different addressing and forwarding assumptions. Early networks didn’t need MAC addresses; they had a single destination per wire. LANs invented layer‑2 addressing to let many machines share a single medium. IPv6 designers made choices to accommodate those realities, and in doing so introduced features and complexity that now look awkward when you compare them to the lean, hacky elegance of IPv4.

So what now?

The takeaway feels almost human: design choices that make perfect sense in one era can feel like baggage in another. IPv6 may have been a “good” design for the world its authors imagined — more structured, less hacky — but the actual Internet is messier, backward-compatible, and full of hard, deployed realities. Can we have elegance without breaking what works? That’s the ongoing question. In the meantime, engineers will keep squabbling at standards meetings, patching, and occasionally discovering something genuinely better — and we’ll all keep using whatever gets the job done.

Sources: apenwarr.ca, Hacker News