Content-Type: text/plain;
	charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 13.0 \(3608.60.0.2.5\))
From: Tim Bruijnzeels <tim@nlnetlabs.nl>
In-Reply-To: <20200512211314.4C2CD20156DBC4@minas-ithil.hactrn.net>
Date: Wed, 13 May 2020 11:22:44 +0200
Cc: sidrops@ietf.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <E4C00A94-B0AF-4FEA-8752-7630792337BA@nlnetlabs.nl>
References: <be9450ba-fe9c-465f-98a2-919772b3b32a.ref@verizon.net>
 <be9450ba-fe9c-465f-98a2-919772b3b32a@verizon.net>
 <20200512211314.4C2CD20156DBC4@minas-ithil.hactrn.net>
To: Rob Austein <sra@hactrn.net>
Archived-At: <https://mailarchive.ietf.org/arch/msg/sidrops/rZHaVcMVZttlkcoJMv5KUGO-02A>
Subject: Re: [Sidrops] rev 4 (corrected CRLDP source changes, thanks to Tim)
Precedence: list

Hi,

> On 12 May 2020, at 23:13, Rob Austein <sra@hactrn.net> wrote:
>=20
> On Fri, 08 May 2020 11:26:22 -0400, Stephen Kent wrote:
> ...
>> A manifest SHOULD contain exactly one CRL (.crl) file and it MUST be
>> at the location specified in the CRLDP in the manifest=E2=80=99s EE
>> certificate.  If more than one .crl file appears in the manifest,
>> only file names matching the CRL specified by the CRLDP will be
>> processed. If more than one .crl entry appears in the manifest, and
>> matches the CRLDP, the first one encountered MUST be used.  Any
>> other .crl files MUST be ignored and a warning MUST be issued.
>=20
> I went back and looked at how my RP code handled this.  One can very
> quickly get lost in the weeds here, but briefly: I start with a set of
> "candidate manifests" and a set of "candidate CRLs", and keep pruning
> those sets down with one form of validity check after another
> (signatures and hashes must match, timestamps must be sane, URIs must
> match, yada yada).  If, at the end of this I still have more than one
> candidate CRL, I don't necessarily pick the first CRL in the manifest:
> instead, I sort the candidates by CRL Number, thisUpdate, and time at
> which I retrieved that CRL object (in descending order of preference,
> so the timestamps only matter if CRL Number is identical, etc), and
> use this to pick the "most recent" valid CRL.
>=20
> YMMV, but this arguably yields a more useful result in this screwball
> situation.  That said, this is (obviously) more complex to describe
> and to implement, so may not be worth it, given that this should never
> be happening in the first place.
>=20
> If one wants a simplified version of this algorithm that stays within
> the confines of a single manifest, one could do the sort by CRL
> Number, then thisUpdate, then position in the manifest.

To the best of my knowledge there has never been any RPKI CA =
implementation which will use more than one .crl per manifest. So, can't =
we just simplify life and make sure that no one does in future?

It won't break any existing code, we just need to make it clear to new =
CA implementers that they should not needlessly complicate life for =
themselves and everybody else.

Tim