[ih] IETF relevance (was Memories of Flag Day?)

[Dave Taht]

I tried in the early days of the bufferbloat effort to bring some
focus in the ietf towards "Running code", over consensus. To this day,
most there seem to think that complex standards, invented out of
theory, can actually be made to work.

The IETF is at its best, when proven algorithms and running code are
brought to it. I shudder to think about all the handwaves that enter
into the standardization process without running code to complement
it. Quic benefited from  3 years of exhaustive testing in the field
before it was brought to the ietf. eBPF, now entering the standards
process there, after 6 years+ in the field, evolving, may well
benefit.

fq_codel - which took a weekend to write in 2012, took 6 years to
standardize as RFC8290.

Exhaustive specifications and wishful thinking that end up impossible
to implement well (not just ipv6, but 3gpp and many other standards
issued in the last two decades) saddle us all with a lot of wasted
work if only the spec and implementation processes proceded more
concurrently, or - code first, spec later.

The coders, long ago, fled the room, screaming. I mostly check in and
participate nowadays to see what the latest insanity is.

I also really wish the ietf acted to revise or retract old standards
that did not work out. One going around recently is the only partially
implemented :0 ipv6 anycast feature which gives :0 networks a
capability that is dubious  at best, and leads to added configuration
complexity: https://www.daryllswer.com/behavioural-differences-of-ipv6-subnet-router-anycast-address-implementations/