Re: [Doh] panel discussion on DoH/DoC

n Sun, Feb 10, 2019 at 4:11 PM Andrew Sullivan <ajs@anvilwalrusden.com>
wrote:

> Also, what do you mean exactly by "picking one server"?  The back end
> state
> surely (!?) won't be shared and there's no reason to suppose names or
> numbers get you to "the same" thing. So ...?
>

The general idea would be to have a list of DoH servers (operators, not
machines) and then to randomly select one for each client. We haven't
decided to do this, so also haven't decided on how we'd implement it :)

-Ekr

> A
> --
> Andrew Sullivan
> Please excuse my clumbsy thums.
>
> On February 10, 2019 18:51:37 Martin Thomson <mt@lowentropy.net> wrote:
>
> > On Sat, Feb 9, 2019, at 08:19, Joseph Lorenzo Hall wrote:
> >> Those are great points and I may have taken some liberty with the the
> >> "multiplexing across many DOH providers" statement as I'm not really
> sure
> >> what browsers want to do, it just seems like having one or a few
> options is
> >> not ideal. Anyway, I recognize that the work I'm interested in is not
> >> protocol work (as far as I can tell), so I'll stop bugging folks here!
> >
> > I don't think that this is entirely crazy.  I agree with Shane that
> trying
> > to load balance between servers or randomly route to servers is likely
> to
> > be worse than just picking one, but one option we've considered is
> picking
> > one server randomly and using that consistently thereafter.  (Cue all
> the
> > second-order questions regarding how to use that for tracking, etc...)
> >
> > There is also talk of finding ways to route requests to servers that
> might
> > own the same name, but that's far more risky.  In other words, this is
> not
> > a decided matter because it is so clearly not simple.
> >
> > _______________________________________________
> > Doh mailing list
> > Doh@ietf.org
> > https://www.ietf.org/mailman/listinfo/doh
>
>
>
> _______________________________________________
> Doh mailing list
> Doh@ietf.org
> https://www.ietf.org/mailman/listinfo/doh
>