Le 20/05/15 15:08, Kiran Ayyagari a écrit : > On Wed, May 20, 2015 at 8:37 PM, Shawn McKinney <[email protected]> > wrote: > >>> On May 20, 2015, at 2:43 AM, Emmanuel Lécharny <[email protected]> >> wrote: >>> >>> interesting result : in the past four days, we went from 2543 Sonar >>> issues down to 1677. No more blockers, only 1 critical instead of 61, >>> and many less minor issues : >>> >>> https://analysis.apache.org/dashboard/index/211987?did=4&period=2 >>> >>> This tool is very intersting in many ways. It also let you know which >>> are the undocumented API in a convenient way : >>> >>> >> https://analysis.apache.org/dashboard/index?id=org.apache.directory.fortress:fortress-core&did=2&period=2 >>> (last table on the right column) >>> >>> Nice ! >> Agreed, very handy tool indeed. Has enabled us to find many potential >> problems and fix rapidly. >> >> My question is how far do we go? For example the code complexity and >> cycle metrics. The former complains about simple utility methods that have >> many getters and setters. The latter requires moving classes around thus >> breaking compatibility with existing clients. I get the need to reduce >> complexity, yet don’t want to spend time needlessly. >> >> I personally prefer to maintain backward compatibility unless it is a > major version release. And cause the > code base is quite large, IMHO, it is "ok" even if this cleanup effort > spans multiple major versions.
Considering Fortress has not reach an official 1.0 yet, I do think it's the last time to change things before we freeze the API.
