My initial reaction is 4b sounds like a good idea. But I don't think we need to "start from scratch" but simply go back to what we had in 0.2, which provides this.
http://incubator.apache.org/deltaspike/security.html On 26 Jul 2012, at 23:49, Romain Manni-Bucau wrote: > great summary! > > of course 4b) is IMO more reasonnable. Then the question should be do we > start from scratch? -> new api, new conf, new impls > > - Romain > > > 2012/7/27 Mark Struberg <[email protected]> > >> Oki, here we go. >> >> We had a quick chat about where we basically stand today. >> >> >> This is not intended to be a a 'what shall be' but more a 'what do we >> have' + 'what do we really need' email. >> >> >> 1.) What we have today: >> I've looked at the Security module and what I understand it's pretty >> powerful and complex. >> There are aprox. 30++ Interfaces which are very flexible but also very >> hard to get right. Having lots of flexibility also makes it easy to do >> things wrong as user. E.g. IdentityManager which allows to create users. >> The RoleQuery and the whole Role management is pretty complete from the API >> level but I've never seen it used in such detail in any application yet. >> Most times there is an additional mapping role -> rights. And the right is >> what gets used in the application (e.g. in rendered= ). >> >> 2.) What is available in projects: >> In my last 10 projects we never had the choice to define our own login >> logic. Some customers had radius, others authenticated against SAP or >> kerberos. Then there are some LDAP and we even have a single sign on based >> on Smalltalk. And there is absolutely no way to get rid of those! Most of >> the time you cannot even create your own users... Of course there is the >> need for a simple html based user login for _some_ applications. But this >> is most times only needed for green-field projects. Whenever you do >> projects for a bigger company you most likely will find some well >> established SSO in place. >> >> 3.) what is needed in those projects: >> I did quite some integration already in the past and the only thing which >> we did really need was >> >> 3.a.) to express some interrest: "current user likes to do actionX" >> This can be done via a @Secured interceptor, via @ViewConfig, via >> @PageBean etc -> might get provided by DS. >> >> 3.b.) to evaluate the "is the current user allowed to do actionX" >> Like with JAAS Voters this can be done via a simple Interface which >> returns a boolean. This is really similar to what Seam2 had and also what >> CODI did. >> All the evaluation and binding to an existing authorisation and >> authentication can be done in this AccessVoter/checkPermission. -> we might >> provide the Interfaces in DS. The impl is _always_ up to the user. >> >> 4.) what are our options: >> >> 4.a.) fully implement our own security manager. This will surely still >> take some time as this is a complex topic! Many of the interfaces are ok >> but there is not yet an impl behind it. My personal estimation is that we >> now hit the 15% line, and a few people already spent a good amount of power >> for it. So this will not be finished for the next 5 months I fear. >> >> 4.b) implement a simple Voter + @Secured and let the user deal with the >> rest. In both Seam2 and CODI this turned out to not only be extremely >> flexible, but it is also rather easy to integrate [1]. We could also >> provide an additional module which contains a composite component with >> login userId + pwd fields + a simple backend for it. But just as a small >> additional module which might optionally be used for easier integration >> into JSF apps if there is not yet an existing SSO implementation. >> >> LieGrue, >> strub >> >> >> [1] >> https://github.com/struberg/lightweightEE/blob/master/gui/src/main/java/de/jaxenter/eesummit/caroline/gui/security/AdminAccessVoter.java#L36 >> >> >> ----- Original Message ----- >>> From: Jason Porter <[email protected]> >>> To: [email protected] >>> Cc: >>> Sent: Thursday, July 26, 2012 9:03 PM >>> Subject: IDM impl feedback >>> >>> T he implementation that's in HEAD right now is incomplete. There are many >>> methods which are basic IDE generated stubs in multiple classes. I'll >> hold >>> off on any feedback until it's complete. >>> >>> -- >>> Jason Porter >>> http://lightguard-jp.blogspot.com >>> http://twitter.com/lightguardjp >>> >>> Software Engineer >>> Open Source Advocate >>> Author of Seam Catch - Next Generation Java Exception Handling >>> >>> PGP key id: 926CCFF5 >>> PGP key available at: keyserver.net, pgp.mit.edu >>> >>
