Hi Stephen,

Ok, I understand the model you talk about now. Yes the CPU may not be
the biggest concern as the server is verifying the Cert's offline. I
guess this would also lead to models like CRL's for revocation.

Like I said earlier as SIDR does not stop malicious attacks, but only
ones caused unintentionally, is it not possible to actually use a
simpler mechanism to get over such errors?

Thanks,
Vishwas

On Thu, Feb 28, 2008 at 3:35 AM, Stephen Kent <[EMAIL PROTECTED]> wrote:
>
>
> At 8:31 PM -0800 2/27/08, Vishwas Manral wrote:
> Hi Stephen,
>
>  The point I raise is that there is a cost associated with this, using
>  certificates has a CPU cost associated with it.
>
>  I may be missing the point but even if you leave aside the cost of an
>  off line server to do this check but if the checks are done on each
>  new prefix we can still overload the off line server. If we do things
>  like rate limiting we can still have an attacks, or cause delays in
>  the convergence times.
>
>  Thanks,
> Vishwas
>
>
> I think you are missing the point. So long as the processing is done as an
> offline operation, not as a gating item in routing, it does not strike me as
> a DoS concern. The initial use of the infrastructure is analogous to
> downloading IRR databases and processing the RPSL assertions, an operation
> many ISPs perform today on a daily basis.
>
>
> More to the point, folks have implemented the necessary software and tested
> it with about 20K certs and CRLs and 10K ROAs, a reasonable starting point.
> I don't have the precise figures in front of me now, but I believe their
> results show that the time to do all the processing (on a laptop) is about
> 20-30 minutes, and the time is dominated by the retrieval of the data from
> online repositories, not by the crypto operations per se. For a once daily,
> offline operations, this seems quite reasonable.
>
>
> Steve
_______________________________________________
Sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr

Reply via email to