front controller pattern and security

[John Gregg <jo...@techarch.com>]

Hi all.

I've been thinking about how the j2ee front controller pattern (used by
Struts et al.) does/does not take advantage of url-based authorization
constraints in web.xml.  I want to avoid having to check roles in my own
code as much as possible.  At first I thought I could declare a URL like
/somewebapp/somerole/* to require the "somerole" role before being allowed
access to my controller servlet.  Another URL would be
/somewebapp/someotherrole/* but would map to the same servlet.  That servlet
would then pick off the action at the end of the URL and execute it.
However, while I can restrict access to the servlet, or whatever other
"physical" resource I'm trying to protect, what I really want to protect is
the action that's executed.  Am I just stuck with enumerating all possible
actions in by web.xml (/somewebapp/somerole/someaction,
/somewebapp/somerole/someotheraction, etc.)?  Should I instead make a filter
that enforces this for me?  I'm facing the same problem with Apache SOAP's
rpcrouter.

thanks

john


--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: front controller pattern and security

[Martin Jacobson <ma...@libero.it>]

John Gregg wrote:

> Hi all.
> 
> I've been thinking about how the j2ee front controller pattern (used by
> Struts et al.) does/does not take advantage of url-based authorization
> constraints in web.xml.  I want to avoid having to check roles in my own
> code as much as possible.  At first I thought I could declare a URL like
> /somewebapp/somerole/* to require the "somerole" role before being allowed
> access to my controller servlet.  Another URL would be
> /somewebapp/someotherrole/* but would map to the same servlet.  That servlet
> would then pick off the action at the end of the URL and execute it.
> However, while I can restrict access to the servlet, or whatever other
> "physical" resource I'm trying to protect, what I really want to protect is
> the action that's executed.  Am I just stuck with enumerating all possible
> actions in by web.xml (/somewebapp/somerole/someaction,
> /somewebapp/somerole/someotheraction, etc.)?  Should I instead make a filter
> that enforces this for me?  I'm facing the same problem with Apache SOAP's
> rpcrouter.
> 


Surely, the way you describe, using servlet-mapping & auth-constraint 
does exactly what you want?
In other words, you have one (controller) servlet declared, and you use 
servlet-mapping to declare all urls that invoke that servlet. In 
auth-contstraint you declare the mapping between urls and user roles. 
So, when a user invokes /app/manager/ac_details, the auth-constraint 
would trigger an authorisation check to make sure the user was in the 
indicated role. Then (if he's allowed), because the servlet-mapping maps 
this url to the controller servlet, the controller servlet is invoked, 
and, as you say, you can then determine what the action was (ac_details) 
by examining the url. This method does not restrict access to the 
controller servlet directly, but does protect the functions to be performed.

Or have I misunderstood?

Martin



--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>

Re: front controller pattern and security

["Craig R. McClanahan" <cr...@apache.org>]

On Thu, 13 Jun 2002, John Gregg wrote:

> Date: Thu, 13 Jun 2002 10:44:15 -0500
> From: John Gregg <jo...@techarch.com>
> Reply-To: Tomcat Users List <to...@jakarta.apache.org>,
>      john.gregg@techarch.com
> To: tomcat-user@jakarta.apache.org
> Subject: front controller pattern and security
>
> Hi all.
>
> I've been thinking about how the j2ee front controller pattern (used by
> Struts et al.) does/does not take advantage of url-based authorization
> constraints in web.xml.  I want to avoid having to check roles in my own
> code as much as possible.  At first I thought I could declare a URL like
> /somewebapp/somerole/* to require the "somerole" role before being allowed
> access to my controller servlet.  Another URL would be
> /somewebapp/someotherrole/* but would map to the same servlet.  That servlet
> would then pick off the action at the end of the URL and execute it.
> However, while I can restrict access to the servlet, or whatever other
> "physical" resource I'm trying to protect, what I really want to protect is
> the action that's executed.  Am I just stuck with enumerating all possible
> actions in by web.xml (/somewebapp/somerole/someaction,
> /somewebapp/somerole/someotheraction, etc.)?  Should I instead make a filter
> that enforces this for me?  I'm facing the same problem with Apache SOAP's
> rpcrouter.
>

A Filter would be a good solution if you are running on a Servlet 2.3
based platform (like Tomcat 4).

Another approach is to see if your controller framework can help you out
-- for example, Struts 1.1 lets you define the roles that are allowed to
access a particular Action with the "roles" attribute on your <action>
element.  The simplest way to use this is to create a security constraint
against URL pattern "*.do" that just forces the login, and then use the
role restrictions on the actions to define which actions can be executed
by which logged-in user.

> thanks
>
> john
>

Craig


--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>