Re: [dnsext] we need help to make names the same, was draft-yao-dnsext-identical-resolution-02 comment

[Nicholas Weaver <[e-mail]>]

On Feb 22, 2011, at 6:47 PM, Masataka Ohta wrote:

> Nicholas Weaver wrote:
> 
>> OR use DNSSEC but sign data dynamically.
>> 
>> 10 years ago, online signatures may have been questionable
>> computationally.  Today, online signatures are near-trivial
>> computationally.
> 
> Any reference?

Not for DNSSEC, but computationally DNSSEC signature generation is no harder than SSL connection setup.  Google was able to make Gmail SSL only with a trivially small increase in computation.


Random reported SSL setups per second are ~1500/s for a commodity 64B processor with 1024B keys.  (People don't benchmark this kind of thing as much as they used to...)

AND this parallelizes obscenely well: because of the stateless nature of DNS, a simple salted hash function dispatch to compute nodes will easily handle a cluster to however many requests per second you need to case-sensitive-normalize.  

AND such compute nodes could cache the signatures for common conventions, which means you can't computationally DOS the common cases and the COMMON cases no longer require dynamically executed DNSSEC...


> Note that "near-trivial computationally" means more lengthy
> signatures are required than 10 years ago.

2048B signatures are only ~4x harder than 1024B signatures computationally to GENERATE, but vastly harder to crack using brute force.  Thats the wonder of cryptography and the exponentials involved.  Its quadratic increase in complexity to add bits to your RSA keysize, but far far far greater impact on the cost of cryptanalysis.


_______________________________________________
dnsext mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/dnsext