[cgiapp] app authorization patterns, best practices?

[Jerry Kaidor]

> I have some questions regarding best practices when implementing
> role based access control (RBAC). I have been playing with
> CApp::Authentication and Authorization, and they both do basically
> what I need.

*** Me too.  I have three sub-businesses.  Let's call them A, B, C.  I
want to have access to all of it, but I want my managers to only have
access to
their particular piece.

    Here's more or less how I did it.  BTW, it recently broke for unknown
reason, and I havent' gotten around to troubleshooting.

   CApp:Authentication has a notion of "drivers".  I wrote a multi-DB
driver that can look in a set of authentication databases.  I have a
"users"
mysql database at my global level, and another "users" database in each
subbusiness.  When somebody attempts to log in, the multi-auth driver
tries each database in turn.  If it successfully authenticates against the
"global" database, that user has privs for all the sub-businesses.

   Each line in the "users" database has a set of "permissions" flags,
which correspond to things seeming to permission as I coded.  In line
with the code are statements in the form if( getpriv( user, business
)){}.  That way, I have a permissions system with much finer
granularity than Capp::Authorization, which I do not use at all.

                               - Jerry Kaidor ( jerry at tr2.com )





>
> Here's the skeleton I came up with -
>
> 	https://gist.github.com/33d23edf8fa2c0f48dc0
>
> My question is really, what's the best way to go about separating
> functionality in a CApp based application?
>
> A practical case I am looking at right now is that I have form that
> is used to manage user data. There are 3 roles - User, Manager, and
> Admin. Each one has the types of permissions you'd expect (User can
> manage himself, Manager can manage his Users, Admin can do anything).
>
> I was thinking of the best way to build this form and control actions
> cleanly and compose this form using 3 different runmodes that are
> increasingly restrictive.
>
> For example, the User form calls the "user" runmode, and returns the
> form content. The "manager" runmode takes the output of "user" and
> adds some stuff to it. The "admin" runmode might take the result of
> the "manager" runmode - which would also include what the "user"
> runmode provides...and so on.
>
> Ultimately, my goal is to get away from nasty frog boiling "if" blocks
> controlling authorization and rely on composable functions (i.e.,
> runmodes or modules) that will cleanly give me what I would like using
> the runmode level protection that CApp::Authorization provides you.
>
> I've searched around and banged my head against this pretty hard, so
> any thoughts or resources would be appreciated. For all I know, this
> might be a bad idea. But I am really just looking for the best way to
> create an access controlled system as cleanly as possible.
>
> Thank you,
> Brett
>
> ps: I noticed that even if POST_LOGIN_RUNMODE is protected via
> CApp::Authentication, the check seems to be ignored immediately after
> login. I am not sure if this is a known issue or that there are some
> callbacks happening in the wrong order. This will happen in the gist I
> linked above.
>
> #####  CGI::Application community mailing list  ################
> ##                                                            ##
> ##  To unsubscribe, or change your message delivery options,  ##
> ##  visit:  http://lists.openlib.org/mailman/listinfo/cgiapp    ##
> ##                                                            ##
> ##  Web archive:   http://lists.openlib.org/pipermail/cgiapp/   ##
> ##  Wiki:          http://cgiapp.erlbaum.net/                 ##
> ##                                                            ##
> ################################################################
>