Re: Example of stupid inconsistencies between registries

On 5/03/2012, at 11:43 PM, Patrik Fältström wrote:

> [snipped]
> Reactions?


If we start with the recognition that some registries want thin DNSSEC while others want thick and this difference is currently irreconcilable then what we are really talking about is the balance between consistency and choice in EPP and whether that's good now and will it get better over time.  Your complaint in essence is that in this case balance is not good now as choice has won out over consistency.  

I tend to agree with you that this is not good enough now because, in my view, policy and protocol have been too tightly coupled. 

To explain what I mean, here's a simple idea - could the protocol have been designed to better minimise pain for registrars while maintaining the choice for registries?  For example, could the interface have required registrars to provide both the DS records and the keys with registries using whichever of the two they want?  The implication of doing so is that we get consistency of interface while individual registries maintain choice on policy.  Is a registrar more or less likely to be confused by this than by two variants of the protocol?  My view is that they have to learn the policy difference anyway but they don't need to learn policy from implementing the protocol.

The second question is whether this balance is going to get better in future, where again I am not that hopeful.

With two variants of the protocol as tightly coupled to policy as this, then when a registry changes policy it has to change interface.  For a start that's a disincentive to change as it imposes a cost on registrars.  There's also the issue that they registry making the change doesn't have the historical data to conduct a 'what if our policy had been different' exercise.

The more general problem is that there is a strong school of thought around EPP that goes
- the core protocol is done and does not need changing
- everything else we need can go in an extension
- competition will sort out what extensions (or variants within extensions) dominate

I would suggest that is insufficient because policy, which eventually drives the technology, works differently and so the core will gradually become less suitable and the cruft of undead extensions (and variants) will become unmanageable.  Instead a more methodical process is needed that periodically reviews the core and extensions and where appropriate change the core and kills extensions.  Some of this review needs to be quite rapid to keep pace with the rapid pace of policy change.

cheers
Jay


-- 
Jay Daley
Chief Executive
.nz Registry Services (New Zealand Domain Name Registry Limited)
desk: +64 4 931 6977
mobile: +64 21 678840

_______________________________________________
provreg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/provreg