At 5:57 PM +0800 2/23/11, George Michaelson wrote:
PKIX is discussing the effect of a missing CRL at the moment.
It appears some people doing relying party tests in 5280
over-interpret the non-availability of a CRL as a hard-fail, and
treat the CA products as invalid, in a belief that 5280 requires
this of them.
But, this is not actually required. And in like sense, it would not
be required of a manifest.
There are of course certain questions regarding signed products
which become unanswerable in the face of not having a manifest, or a
CRL, but they are not hard-fails which would invalidate the entire
object set, and by extension, the sub-tree of child products ad
infinitum.
section 6. Local Cache Maintenance of the ARCH document does place
an onus on a cache administrator to review the CRL/Manifest products
and detect corruption, but its not presented as a strong/normative
validation issue.
Can I ask relying party tool writers to consider this in their
tools? Warnings are useful, but over-aggressive pruning or
de-validation would not be good.
cheers
-George
The BBN RP software embodies the notion of stale (vs. expired) CRLs
and manifests. Mark and Andrew are authoritative on this matter, but
my recollection is that we give the RP knobs to control how they
choose to deal with stale items, maybe even based on te degree of
staleness.
Steve
_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr