Re: Tussles in IPv6 Land

[james woodyatt <[e-mail]>]

On Jun 12, 2017, at 14:08, Brian E Carpenter <[e-mail]> wrote:
> 
> And so we've morphed this thred into another thread about /64...

I’d rather not. My point is complementary to Lorenzo’s point in <[e-mail]>, where he implored the group to remember that the other side of the tussle involving operators and their needs to manage their networks, i.e. the network engineers, is the side comprising the users of the network and the designers of human interfaces for machines and applications that communicate via the network, i.e. the host engineers.

All three of the major tussles you mentioned in the message that kicked off this thread involve tensions between these two groups with often conflicting interests. It’s not just the /64 issue.

The DNS server address configuration issue is contentious because different human interface designers are concerned with different problems than network engineers, and we therefore have two methods of configuring host operating systems because operating systems are designed by different developers with different objectives in mind. The tussle between network engineers and host engineers here is about which side must carry more of the burden of ensuring interoperability between hosts and the network.

Likewise, from my perspective (which is quite far away from the heat of the action), the tussle between network engineers and host engineers with regard to header insertion seems to revolve around the effect of header insertion on path MTU discovery, which has implications for network application design. I, for one, am skeptical about plans to admit header insertion not because I have some commitment to theoretical purity, but precisely because I’m concerned about ever, some day, in our glorious future, having any kind of decent expectation that path MTU can be resolved over the Internet with any reliability, because that has serious implications for application design.

The point I wanted to add over and beyond the one that Lorenzo was making is that we are not all just network engineers here. Enumerating a list of all the possible types of network operator and putting forward the view that all the important tussles in IPv6 Land are between operators of different kinds of networks, neatly and conveniently erases a whole other class of people with networks for which there are no operators at all, a class of people who number orders of magnitude more than the operators, and who have a stake in the tussle represented by a whole other category of engineer, the host engineers, whose interests diverge more significantly from the interests shared among all the disparate subclasses of network engineer.

Maybe erasing them from the discussion was an accident, and my point was to try to correct what I hoped was merely an oversight and not an active deliberate effort to shut them out.


--james woodyatt <[e-mail]>