"Phillip Hallam-Baker" <[EMAIL PROTECTED]> writes:
>I think you are probably refering to Ron's paper in FC'98. I presented an
>alternative and somewhat radical architecture at RSA'99 which demonstrated
>that it was practical to distribute revocation info in real time for a
>population of 5 billion certs.
There are many good alternatives (actually pretty much everything is better
than CRL's, so it's difficult to come up with a bad alternative), but the
problem they all have is that they're not CRL's. To paraphrase Bob Jueneman
"The market has spoken. The answer is CRL's, although noone can quite remember
what the question was". Given that it's going to be very difficult to make any
headway against this unless you've got a vertical-market application where you
can design things the way you want them, my approach has been to try to turn
CRL's into a silk purse through some form of reprocessing (a CRL -> OCSP
gateway would be an example of this). That way, you can pretend to have CRL's
(giving the customer exactly what they asked for) while also having a system
which works. The warning from Padlipsky's "Elements of Networking Style" is
still appropriate here though for anyone trying to work around the problem of
CRL's: "The schoolmen couldn't find how many teeth a horse had in Aristotle; a
student suggested they look in some horses mouths. They expelled him".
Peter.
