Re: [Doh] GDPR and DoH

[Stephen Farrell <[e-mail]>]

Hiya,

On 07/04/2019 17:28, Jim Reid wrote:
> On 7 Apr 2019, at 15:07, Stephen Farrell <[e-mail]> 
> wrote:
>> 
>> From the above, it sounds like you agree there is no effect on the 
>> DoH WG.
> 
> I didn’t say that Stephen.

Sure, but I figured it might follow from what you did say.
(That said, your mail to Christian clarifies a bit for me
where we may disagree here.)

> 
> IMO GDPR has no impact on the DoH *protocol*.

Agreed.

> GDPR does have an impact on how that protocol gets used and deployed.

True, but I disagree:-)

Where I disagree is I don't think it makes any sense to try
analyse the privacy or GDPR issues of DoH in isolation (that's
sort-of been done and the DoH-specific issues only relate to
the HTTP framing and how that provides more tracking tooling if
a service provider wants to try use those and if a client isn't
a bit more strict than a typical web UA). So those DoH-specific
issues are really just a footnote to the main concerns which
mostly relate to service selection, centralisation and perhaps
transparency.

OTOH, it makes total sense to do such analysis for the DNS more
generally, perhaps splitting out the stub/recursive scenario as was
done when we started DPRIVE.

> This needs to be discussed at the IETF and the obvious home for that
> discussion seems to be the DoH WG - at least for now. The WG is
> explicitly chartered to work on DoH’s security and privacy issues. If
> not there, then where?

That's one for the IESG, and after Prague they do know for sure
that that's one for them. I'm hoping they do that bit of steering
soonish.

> 
> You said you "agree that we need to consider GDPR issues, where they 
> exist” and seem to be saying that there are no new GDPR issues
> caused by DoT/DoH. Well, the issues are new. Sort of. Although
> they’ve been around for a while, as you said they “have been largely
> or totally ignored until very recently”. Maybe they should have been
> discussed earlier. But that earlier non-discussion doesn’t seem to me
> to be a valid reason for dismissing these concerns now that they have
> been raised. YMMV.

Mostly agree, with two caveats:

- I think the newness here is really driven by other concerns
with the kind of change exemplified by the mozilla/CF setup, and
is not solely/primarily driven by concerns with privacy/GDPR,
where I continue to maintain that there aren't really new issues.

- I am not "dismissing" privacy/GDPR concerns - I'm claiming that
they are a) not new b) not specific to DoH c) equally apply to
services using Do53 as well as DoT and lastly d) are only worth
analysing in that broader context.

Bottom line is I think we're mostly in agreement about the meat
of the issues here, but seem to still disagree about the best
venue and scope for discussion. And that last bit is the IESG's
to fix, so we don't have to:-)

Cheers,
S.

> 
>