Re: Link-local [was Re: draft-ietf-6man-slaac-renum: Processing of PIO Lifetimes at Hosts]

[Ted Lemon <[e-mail]>]

On Sep 10, 2020, at 4:19 AM, Nick Hilliard <[e-mail]> wrote:
> Apart from the mind-boggling disparity between the two positions you're suggesting here, I'm pretty disturbed by Philip's email because of two things: firstly how inherently complicated ipv6 is, and secondly this is ietf 6man and like Philip, I'll bet that most people in this WG wouldn't be able to answer all 4 questions correctly.  I know I couldn't.
> 
> This poses the question: is 6man an appropriate working group to be entrusted with the maintenance of the ivp6 protocol?

There’s also v6ops.

I think that the working group chairs hold a strong editorial position, and this is arguably not ideal, but on the other hand, if they did not, I suspect that we’d do a lot more harm. If I were to suggest an improvement, it would be that this were done more openly, with more transparency, and with a somewhat muted assumption that everything is already perfectly correct, which I do sometimes notice (sorry, guys!).

That said, any networking stack is going to be complicated. IPv4 is differently complicated. Are you an expert on the many different kinds of network address translation, how they succeed, how they fail, and how to write an application that will win in the maximal set of cases? How about DNS? Do you know all the ways that middleboxes can break DNS, and how to win when they do? Do you know how to correct a bufferbloat problem? Do you know how to configure a multi-subnet home network so that devices on all subnets can discover and reach each other?

The answer is likely that you do not. All of these, including typing in IPv6 addresses, are things that experts know, not things that end users know. End users type in names if they type in anything. They click on URLs. They plug their routers together using similar-looking cables. They do not type in IP addresses.

So the set of bugs I pointed out in the Linux stack need to be fixed, for sure, but the basic issue comes down to good UI design, not to making every user an expert in our obscure field.