[Sidrops] Re: Call for adoption: draft-spaghetti-sidrops-rpki-erik-protocol-04 (Ends 2025-11-29)

[Tim Bruijnzeels <[e-mail]>]

Hi Martin,

The thinking among co-authors, or at least this co-author, is that we will
likely want to deprecate at least one of RSYNC or RRDP if and when this is
done.

But it is too early to have an informed discussion about that until we adopt
this work, do proof of concept implementations, and learn.

In the early stages at least the erik relays provide and additional and
fully optional way to get the data. We believe that this is useful in and
by itself, because it does not necessarily require repository operators to
make any changes. Nor are CA implementations required to make any changes.
The RRDP and rsync URIs remain unaffected, but Erik relays can act as
additional caches. This provides mirroring, redundancy and better availability
for hard to reach repos that we see today. For a repo like RIPE NCC's this
can help as it will reduce data churn.

But sure, if and when we feel this works well enough to make it mandatory to
implement, we should also have that discussion about deprecating something else.
Then we may also need to update the certificate profile, signed object profile,
provisioning and publication.

We can have that discussion in future, but again, I think that now we are not
yet informed enough.

Tim



> On 10 Nov 2025, at 14:27, Martin Hoffmann <[e-mail]> wrote:
> 
> Russ Housley via Datatracker wrote:
>> Subject: Call for adoption:
>> draft-spaghetti-sidrops-rpki-erik-protocol-04 (Ends 2025-11-29)
>> 
>> This message starts a 3-week Call for Adoption for this document.
> 
> I want to support adoption of this document, but at this point I am
> missing an indication -- both by the authors and the working group --
> as to what the aim of defining a new RPKI transport protocol is.
> 
> Specifically, I want to avoid that at the end of the process we have
> three protocols that all need to be implemented by publication server
> and relying party.
> 
> Ideally, I would want the outcome to be that the new protocol is the
> only required transport protocol and both rsync and RRDP will go away.
> But at the very least, I feel that rsync needs to go.
> 
>  -- Martin