On 8-okt-2006, at 21:46, Cat Okita wrote:
I'd suggest that we refer to a minimum standard (probably as set in one of the related documents such as RFC4055), and follow the same, standard means of specifying any 'upwards' improvements in algorithm/key length.
The current work on certificates assumes that these are kept outside the actual routing system, but proposals such as S-BGP carry certificates and signatures in BGP. This will increase the amount of data that must be communicated from one BGP speaker to another, and, more importantly, the amount of data a BGP router must keep in memory manifold. As such, it would be very helpful to keep all of this as small as possible, and use better algorithms with shorter keys rather than weaker algorithms with longer keys.
It would also be useful to consider how much protection is sufficient. Obviously the security should be reasonably strong, but we may be overshooting the mark if we go for the absolutely strongest crypto we have available.
It might be useful approach to carry medium-strength certificates and signatures inside the routing system and have cryptographically stronger certificates and signatures available out-of-band.
In any event, I would rather give operators the choice of algorithm/ key length strength rather than mandate a certain minimum.
_______________________________________________ Sidr mailing list [email protected] https://www1.ietf.org/mailman/listinfo/sidr
