Re: [Secdispatch] GNU Name System

[Christian Grothoff <[e-mail]>]

Hi! First of all, thanks for reading the draft!

On 7/21/20 11:56 PM, Michael Richardson wrote:
> The document is a bit rough, and could definitely benefit from more
> implementer feedback, but that's of course, why you are here :-)

Indeed.

> There are probably some algorithm agility issues, but I suspect that they
> aren't worse that for ICANN DNSSEC, because the decision to change algorithm
> types can be a partially local decision.

This is actually an area where we already had some feedback via other
channels and are planning on
1) fixing some agility issues, and
2) improving the explanation and write-up to make it
   easier to see how agility can be achieved *and*
   actually give a 2nd instantiation.

Alas, this will require significant changes to the draft and
implementation(s).

> DHT just sort of pops up in the section 3 description.
> And then it isn't until section 6 that we learn that there is a DHT.

Yes, we were already told to improve this part of the write-up. We'll
try, specific suggestions welcome.

> I thought I'd learn more in section 8, but I didn't.
> And then section 9.4 says it's out of scope?

The exact specification, yes. It's way too much for one draft to include
it IMO -- not to mention that GNS can be instantiated over virtually any
DHT.

> I didn't quite figure out how it was used at all actually.
> Since all names are local... wouldn't I always know how to find my
> relationships?

Suppose you have alice.bob.dave.carol.  You must have 'carol' locally,
but you would (typically) lookup 'dave' in carol's namespace in the DHT
already. Basically, the root is local, but the rest is resolved via the DHT.

> I see that GNUnet is being funded by nlnet and NGI, and this is rather good.
> I don't see a clear connection to FSF, and I wonder if they will be happy
> about the naming.  I.e. do you have RMS' blessing here?
> Better to get the name right early.

We do indeed have RMS's blessing.

> If this work is worth doing at the IETF, it probably needs it's own WG.
> There are some RGs that deal with far more distributed networking than the
> IETF has been interested in yet, so maybe that's a route.

Well, we hope to find out what the best home at IETF would be.

> How would it get deploy as other than a niche effort of geeks?

Our plan is to initially integrate it with applications that are
underserved by DNS, like decentralized communication platforms / social
networking applications, identity management (see reclaim:ID) and IoT
applications.

> Do you think that it will be possible to build a GNS Service for Android, or
> for iOS?

For Google and Apple, certainly ;-).  And Apps could certainly link in a
GNS stack for name resolution bypassing the OS resolver entirely if desired.