Brian E Carpenter wrote on 26/10/2019 00:02:
Progress will only come as more and more people stop putting IPv6 in
the "too hard" basket.
maybe it is though? Maybe we underestimate the level of overall
complexity because when we look at any individual component, we can
always explain it away because it's only more complicated by a smidgen.
So for example, we mandate /64 for CPE/residential access and tell
people that assigning lots of /64s is good because it gives people
flexibility, although we don't suggest how to push them further down
into the network or provide an easy way to abstract away all the
complexities of running multiple networks.
We say ULA is fine for local stuff, but no NAT please (this is ipv6
after all), and then we write a 16 page document to tell people how to
select a suitable prefix, and then say it's really not that complicated
because the actual algorithm is only a 6-point sample idea and people
can do their own thing anyway.
We tell vendors that they must implement SLAAC and they don't need DHCP
but by the same token tell them that if they want anything more than
getting a host up and running, SLAAC won't do it, so they look at DHCP
(e.g. DOCSIS, residential DSL, etc) and force the vendors to use both
because we block the 2-3 constructs which would allow DHCP to operate as
a standalone protocol and do the whole lot in a significantly simpler way.
For years we never stepped in when people claimed that ipv6 was better
than ipv4 because it was designed to be easy to renumber, and now we're
here wondering why it's 2019 and there's no way to initiate a
renumbering process for SLAAC, and we frown because tech support desks
recommend disabling ipv6 because it's easier than fixing the underlying
problem because there is no underlying fix because SLAAC can't handle
the situation where the CPE renumbers but the end host doesn't.
This is just some bits of residential / cpe access. The same story is
reflected across other ipv6 deployment scenarios.
All of these things are individually reasonable and justifiable, and we
all buy into the explanations because we're good at convincing ourselves
about the things we already want to believe.
But we've crippled ourselves with complexity and we don't want to
acknowledge it.
Nick
