Closed stouset closed 3 years ago
At the time this was written, every paper on the issue recommended p=1. I've been reluctant to let people mess with such things given the amount of _salt_do_not_provide I'm finding hardcoded in people's code. That said, I appreciate sentiment on the p value is under review, so I'll look into this. It permeates down to the C bindings so there's a bit involved.
Thanks for the response. I only noticed the README
comment about parallelism being unexposed as a design decision after opening this ticket, and I can accept that there's some sharper edges around that cost parameter than might be expected.
Perhaps change the salt parameter to _salt_i_hereby_assert_under_penalty_of_international_embarrassment_that_i_am_a_qualified_cryptographer
? :)
Perhaps change the salt parameter to
_salt_i_hereby_assert_under_penalty_of_international_embarrassment_that_i_am_a_qualified_cryptographer
? :)
I can almost guarantee I'll get a PR saying "it's taking me too long to type so I've changed it"
This is now pushed to master as an option.
Right now, parallelism is hardcoded to 1. It would be great if this could be specified along with t_cost and m_cost.