Re: [DNSOP] [v6ops] New Version Notification for draft-v6ops-xie-network-happyeyeballs-00.txt

On Wed., 26 Sep. 2018, 16:10 Ole Troan, <otroan@employees.org> wrote:

> Davey,
>
> If we’re discussing host based versus network based happy eyeballs, would
> it be naive to think that the network based HE would interfere with the
> client’s HE?
>
> A router knows very little about end to end properties of a connection. It
> could of course do those measurements by looking deeply into packets, but
> it would still be restricted to it’s own topological location. Compare that
> with the data available to an MP-TCP host stack.
>
> And note I think HE can’t just be between v4 and v6, but between all the
> candidate connections between source and destination.
>

Yes, multipathing (+ encryption) has a lot of impacts on what is possible
to do and commonly done on the network today. A lot of those activities
would have to be shifted to the host.

Start at slide 63 -

http://www.users.on.net/~markachy/The_Rapid_Rise_of_the_MMHH.pdf

I can't seem to find it now, however I think Fred Baker wrote a draft a
number of years ago that observed that multi-addressing on a single
interface is also a form of multi-homing, meaning that it isn't a
requirement for a "multi-homed" host to have multiple interfaces attached
to multiple and different networks. That means most IPv4 hosts, and all
IPv6 hosts, by design, are potential multi-homed hosts.

Regards,
Mark.

> Cheers,
> Ole
>
> On 26 Sep 2018, at 04:49, Davey Song <songlinjian@gmail.com> wrote:
>
> But in the general case the network cannot.
>> Think host multi-homing.
>>
>
> Yes or no.
>
> Generally speaking the races of IPv6 and IPv4 connections on both network
> and client are going to be suffered by netowrk dynamics, including
> Multi-homing,  route flaps, roaming, or other network falilures. Extremely,
> a client can get a better IPv6 connection in one second (when IPv6 win the
> race), and lose it in next second. In such case, more sophisticated
> measurement should be done(on client or network) , for a longer period, on
> statistics of RTT and Failure rate, or combinations of them. But in IMHO,
> the assumption of HE is relatively stable network for short exchange
> connections. The dynamics exits but relatively rare or no notable impact on
> HE. So I see no such discussion in RFC8035.
>
> Davey
>
> _______________________________________________
> v6ops mailing list
> v6ops@ietf.org
> https://www.ietf.org/mailman/listinfo/v6ops
>