Re: [Sidrops] adopt draft-ymbk-sidrops-rpki-has-no-identity please

[Ben Maddison <[e-mail]>]

On 03/24, Mikael Abrahamsson wrote:
> On Tue, 23 Mar 2021, Randy Bush wrote:
> 
> > from the sec cons
> > 
> >    When a document is signed with the private key associated with a
> >    RPKI certificate, the signer is speaking for the INRs, the IP
> >    address space and Autonomous System (AS) numbers, in the
> >    certificate.  This is not an identity; this is an authorization.
> 
> Agreed.
> 
> Are you opposing the use of RPKI for signing LOAs, because I don't see the
> document affecting the use of RPKI for RSC for signing LOAs.
> 
> The document says it doesn't prove identity. Correct. We all seem to agree
> on that. Now what?
> 
Yup, I certainly agree with that.

I also agree that the LOA use-case for RTA/RSC is fundamentally no
different from the semantics of 6811.

However, "strongly defined" objects (ROAs, GBRs, ASPAs, etc) ship with
well specified semantics, which can be (are) sanity checked by the
originating WG during standardisation.

The difference, for me, and why I think that a document spelling out
what is and is not a suitable use of the RPKI is a good idea, is that
RTA/RSC-style objects give issuers free reign to invent new semantics on
the fly, including ones with bogus trust semantics (like Randy's EFT
authorization).

People will always be able to trip over their own feet using tools like
these, but spelling out what is and is not a good idea seems to be a
useful community service.

I don't think that the document currently does a particularly good job of
articulating this distinction. But I do think that it should be adopted
so that the shortcomings can be fixed by the WG.

Cheers,

Ben