search

FreeAndFair / ColoradoRLA

Software to facilitate risk-limiting audits at the state level, developed for the state of Colorado.
Other
21 stars 8 forks source link

Clarify spec for random seed entry (follow Rule 25? or Starks' tool? or something else?) #625

Closed sfsinger19103 closed 7 years ago

sfsinger19103 commented 7 years ago

A seed of only 6 digits was entered, which generated the message "Please verify that the seed is a numeral at least 20 digits long". Dr. Stark’s tool was able to handle a seed up to 20 digits (20 digits maximum), so why is 20 the minimum for the RLA tool?

I interpret this as a request to allow any integer up to 10^20-1. I'll confirm with CDOS.

dmzimmerman commented 7 years ago

Their rule explicitly says "at least 20 digits", so unless they're rewriting their own rule, this is not something that's reasonable for us to do.

dmzimmerman commented 7 years ago

(this is also a good answer to the question "why is 20 the minimum for the RLA tool?")

sfsinger19103 commented 7 years ago

I'll ask them, but our client is CDOS, not the law.

And there is an equivalence between strings of exactly 20 digits and integers n with 0 <= n <= 10^20 - 1.

dmzimmerman commented 7 years ago

The random seeds are not numbers; they're strings. It's fine for us to change it, but let's make sure we understand that "00000000000000000000" is not equivalent to "0" as a random seed.

sfsinger19103 commented 7 years ago

Yes, indeed they are. Are you suggesting that we deny our client their desired shortcut for entry of strings of 20 digits? I think we should cheerfully add those leading zeroes to whatever they enter.

dmzimmerman commented 7 years ago

Oh? OK. How many of them? How do we know they didn't mean 21 "0"s, or 19, instead of 20? What if they enter 21 0s, do we shorten it to 20? What if they enter 21 1s?

But whatever. Let them actually tell us what they want, and then complain; since they never told us, we made the quite reasonable assumption that they would want what they wrote in their rules. If we limit them to 20 digits, we are explicitly preventing them from doing what their rules say, so that seems infeasible. If we don't limit them to 20 digits, we have no principled way to derive a seed from whatever they enter.

dmzimmerman commented 7 years ago

Your "interpretation", above, is exactly the kind of fuzziness that causes this sort of situation in the first place. So yes, maybe in the future get a clarification before filing this sort of issue.

sfsinger19103 commented 7 years ago

My interpretation isn't fuzzy. It was based on the assumption that they needed a string of exactly 20 digits -- which, as you just pointed out, is not what Rule 25 says.

sfsinger19103 commented 7 years ago

Asked CDOS by email whether they want to change the spec for the random seed from the spec in Rule 25 to something else.

sfsinger19103 commented 7 years ago

Per 9/5 conversation with CDOS, leave this as it was in 1.0.0 alpha 1 -- i.e., follow Rule 25.