Ted and Toke have many emails about DNS and servers and the like. I think that it would be useful to take a few steps back to understand some assumptions. Many of these are tied up in layer 8 and layer 9 issues.
Toke wrote: in my part of the world ISP DNS servers are broken *by design* Ted meanwhile has written that that there are potential situations where one provider may intentially misconfigure something so that a customer will pick a more expensive way to get data. In both cases we are trying to build into our protocols some kind of defense against stupidity. In Homenet, we are really trying to defend against ISP stupidity, while trying to leverage intelligence from the ISP to guard against end-user stupidity. I don't pretend to capture their debate very well. A lot of the DNS publish/public/private, etc. debate centers around tuscles between clue-ful users (engineers) on this list who want to code something sensible so that they can share their clue with the clueless, vs other engineers who want to enable clue-ful ISPs to provide service to clueless users. Then the response from the first group is that this might force them to do stupid things which they have fought against for ages. I can not really characterise the debate very well, but this is what I think I know: 1) ISPs that insist on controlling too much in the homenet will find themselves fighting against Googles, Apples (and Microsoft) who want it to work right. We see this in the CAPPORT debates. By and large, "we" (the public) have "routed around" ISP stupidity, but the collatoral damage to the infrastructure has been immense. (e.g: NAPT44, SCTP, ...) 2) It's very hard to get people to vote with their wallet when it comes to poor ISP performance. Performance usually means speed, but of course there is a whole lot of other variables. The 300Mb/s-marketing-speed fiber vs high-quality 50Mb/s IPv6, and my very technical friends pick the "faster" connection.... Many times, it is because nobody has any actual information. Such as, how often they are really getting 300Mb/s, or how often they are using IPv6 via a bad tunnel. 3) We've tried in HOMENET to automate things; and at the DNS space we are getting to limit of how much policy we can derive automatically. There is SIGNIFICANT work that we can do to automate the implementation of the policy, if we can figure out what the policy is. So I suggest that in order to make progress, we should split things up into ways to implement a policy (assuming the policy is known a-priori), and ways in which can we collect information about when the policy is failing so that things can get fixed; perhaps by a local geek, perhaps by a clueful ISP helpdesk, or eventually, perhaps by a smarter HNCP end-point. -- Michael Richardson <mcr+i...@sandelman.ca>, Sandelman Software Works -= IPv6 IoT consulting =-
signature.asc
Description: PGP signature
_______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet