Le mercredi 9 mai 2012 14:05:47 UTC+2, Ralph Holz (TUM) a écrit :
> Hi,
>
> We're trying to find out the reason again why DigiNotar, when they revoked their certs, inserted an invalidity date that was 1 month later than the revocation date. I seem to recall this was discussed in this group; I'd much appreciate if someone could help me out here.
1 month *later*? Are you sure?
1 month *earlier* would be OK, and the only way to correctly do it.
When producing a CRL, you cannot have a certificate with a revocation date earlier than the lastUpdate field of a previous CRL.
Imagine a CA producing a CRL at date D1, and the certificate X is not present in this CRL.
At date D2 (D2 > D1), you learn than certificate X was compromised at date D0 (D0 < D1).
If you produce a CRL at date D3 (D3 > D2), you cannot declare certificate X as revoked at date D0 (since D0 < D1). You can only declare X as revoked at D2, and add an extension (invalidityDate) specifying D0.
This is imposed by X.509. Paragraph 8.5.2.4 (Invalidity date extension), note 2.