Re: [homenet] use of MPvD in homenet

[Ted Lemon <[e-mail]>]

El 12 ag 2017, a les 20:09, Michael Richardson <[e-mail]> va escriure:
> This seems a bit like FUD: the Internet depends upon all sorts of "brittle" things
> like "not deliberately or accidentially" setting up the routing wrong.

I get that, and it's possible that MPvD is the wrong answer here.   However, I think we need to actually analyze the problem and not just theorize about it at this level.   The reason that the MPvD work happened was that there was a need for it; it may not have been the right way to solve the problem, but I'm having trouble figuring out how LISP or Shim6 improves the situation.

The problem is simply that it needs to be the case that if a host looks up a name from one ISP's name servers, then it needs to use that ISP's prefix to contact the address that it gets from the lookup; if it uses the other ISP's name servers, then it needs to use that ISP's prefix.   Otherwise CDNs fail.

The other problems that you are talking about are also worth thinking about, but we haven't looked at them in this amount of detail.

But one thing to remember is that there's a reason why it's beneficial for hosts to implement MPvD; this is handset vendors implement it it.   It neatly wraps up real problems that real hosts have, pretty much orthogonally to whether IPv4 or IPv6 (or both) are in use.   Given that this is the case, part of the benefit of using it in the homenet is that we'd be using existing logic in hosts.

But if you really think Shim6 or LISP solve the problem better, one exercise that you might want to try is answering the question, "how does Shim6 help me with the CDN problem?   How does LISP help me with the CDN problem.   What do I have to put in a homenet router to make this work?   Which homenet router do I put it in?"   When I consider those questions, I don't know the answers to them.   I do know the answer for MPvD.   So that's why I proposed that approach.