DefaultSecurityManager represents the legacy inheritance-based approach where as ApplicationSecurityManager represents the new composition-based approach. By re-naming the security manager, as you've initially done in v2, developers will be less likely to assume which approach is taken. In other words for your option #2, are you saying that you intend to just rename ApplicationSecurityManager to DefaultSecurityManager, or actually commingle code?
On Sat, Aug 1, 2015 at 7:49 PM, Les Hazlewood <[email protected]> wrote: > Yeah, the idea for deprecating it is that the current > DefaultSecurityManager implementation suffers from excessive subclassing, > as did other Shiro components in 1.x. Shiro 2.x favors much more the > cleaner OO approach of 'composition over inheritance' where functionality > is delegated to components rather than relying on subclasses (and making > things more pluggable / flexible in the process). > > ApplicationSecurityManager is really just a lateral move of the same > behavior of DefaultSecurityManager, but in a different class name so > dramatic class hierarchy changes don't break people currently compiling > against the DefaultSecurityManager hierarchy. > > So there are two approaches for a 2.0.0 final release that I'm thinking > about. @Deprecate DefaultSecurityManager now to ensure people don't use it > anymore and then: > > 1. delete it permanently, or > 2. copy the ApplicationSecurityManager logic into DefaultSecurityManager > at the last minute right before 2.0.0 (making DSM backwards incompatible) > and then deleting ApplicationSecurityManager. > > Both approaches are backwards incompatible, but I prefer #2 just because > DefaultSecurityManager is well-known enough such that not having it might > have more problems than the compiler errors from having different behavior. > > I'm trying really hard to keep all things backwards compatible, as I think > that is a worthwhile goal in a well-established project like Shiro, but > sometimes 'cleaning house' does more good for the community moving forward > than having weird / less-well-designed things stick around and potentially > cause confusion. > > Thoughts? > > Cheers, > > Les > > On Sat, Aug 1, 2015 at 2:11 PM, Darin Gordon <[email protected]> wrote: > > > Les > > > > Would you please confirm that the DefaultSecurityManager is to be > > deprecated as of 2.0 , given the ApplicationSecurityManager ? the > > DefaultSecurityManager hasn't been marked @deprecated yet and so I wanted > > to confirm > > > > > > thanks > > > > DG > > >
