Re: [Add] My single use case

[Martin Thomson <[e-mail]>]

On Fri, Sep 11, 2020, at 01:31, Chris Box (BT) wrote:
> Hi Martin,
> 
> Appreciate you may not be able to answer this until tomorrow but I'd 
> like to map your statement to sections of the draft.
> 
> I think this case is a simpler version of section 3.1 (both 3.1.1 and 
> 3.1.2). I presume you are excluding VPNs (3.3) here, since they are 
> networks in which you have a prior relationship.

Yes, that excludes VPNs and authenticated configuration channels.  It also excludes the "alternatives" thing the draft currently describes in Section 3.

I should have said that this was the thing I want to do *first*.  I think that we'll learn a few things from doing this that will help with those other cases.  I also don't expect people to forget the secondary use cases that they care about and to use those to exert pressure on the design where we are not already fully constrained (that is, if we have to choose between two options where one might be advantageous to a different use case, that is a good way to break a tie).
 
> Personally I would say part of "the matrix" of this use case should 
> involve section 4.1's learning about the scope of names that are 
> relevant.

I don't want to include that in the first version.  My thinking is that this sort of "scoping" is something that can be layered on.

When you join a network, you need a DNS server.  That's fairly simple.  The scoping you talk about is really only necessary if you are really joining two or more networks.  So this reduces to a problem that the provisioning domains and multiple interfaces work already deals with to some extent.  

Dealing with the idea of scoping at that conceptual level is probably more fruitful as it allows our solutions to be simpler (we don't have to muddy the process of choosing to use DoT/DoH with complex questions about whether a resolver is suitable based on what scopes it serves, etc..).  It also allows the overall system to be more robust because we can build fewer, more specific, and more principled scoping mechanisms, as I believe PvDs are intended to be.

I haven't had enough sleep, so that might not be coherent.