[cap-talk] How many bits for a SwissNumber?

[David Wagner]

Tyler Close <tyler.close at gmail.com> writes:
>On 11/29/05, Ben Laurie <ben at algroup.co.uk> wrote:
>> Tyler Close wrote:
>> > Assuming a minimum request time of 10 ms, we would not
>> > expect to guess a 40 bit SwissNumber before 2180, a lot later than
>> > 2030. A 40 bit secret can be encoded in 8 base32 characters. That
>> > seems like a really small secret, but the math seems correct. Am I
>> > missing something?
>>
>> Parallelisation. There's no need to wait for one request before trying
>> the next.
>>
>> Botnets. There's no need to restrict oneself to a single machine.
>
>With the 10 ms time, I was trying to place a lower bound on the
>minimum time the server required to process a request. Are you saying
>that 10 ms is too high an estimate?

Why would it take 10 ms?  10 ms sounds like the time for a network
round trip, or a hard disk seek -- but you can't count network round
trip times for this purposes (due to parallelisation), and I don't
see any reason why the end server would need to do a disk seek to
validate a SwissNumber.  So, what is the server doing to process a
SwissNumber request and determine whether it is valid?  If it is just
computing a hash, then a really fast server might be able to process
as many as millions a second (though I doubt the real number will be
anywhere near that high).  If the server is doing something more
complicated, I don't know what to expect.