[jboss-jira] [JBoss JIRA] (ELY-779) Iteration count transformation: BCrypt

[ https://issues.jboss.org/browse/ELY-779?page=com.atlassian.jira.plugin.sy... ] David Lloyd edited comment on ELY-779 at 11/22/16 6:00 PM: ----------------------------------------------------------- One thing we need to talk about - again - is that BCrypt uses a "cost" parameter which is the log2 of the actual iteration count. I think that we probably would want to at least consider using a real iteration count algorithm parameter to represent the actual count, even though the values have an unusual domain of validity (‹2⁴, ..., 2³¹›, which can be verified by the simple expression {{Integer.bitCount( n ) == 1}}). One challenge of the cost parameter in the implementation is that to raise the cost to the next valid value, you have to add 2ⁿ - 2ⁿ⁻¹ iterations which is hard to fit in with the scheme. We'd probably have to convert to iterations/rounds internally anyway so we can calculate the difference and apply the additional rounds. was (Author: dmlloyd): One thing we need to talk about - again - is that BCrypt uses a "cost" parameter which is the log2 of the actual iteration count. I think that we probably would want to at least consider using a real iteration count algorithm parameter to represent the actual count, even though the values have an unusual domain of validity (‹2⁴, ..., 2³¹›, which can be verified by the simple expression {{Integer.bitCount(n) == 1}}). One challenge of the cost parameter in the implementation is that to raise the cost to the next valid value, you have to add 2ⁿ - 2ⁿ⁻¹ iterations which is hard to fit in with the scheme. We'd probably have to convert to iterations/rounds internally anyway so we can calculate the difference and apply the additional rounds. -- This message was sent by Atlassian JIRA (v7.2.3#72005)