org.forgerock / org.forgerock.openig

15 Sep, 2014

3 commits

  • fc647624   git-svn-id: https://svn.forgerock.org/openig/trunk@576 dbb9e58e-28e6-4ce0-90e8-f11d9605b710 Browse Dir »
    matthew
     
  • 18535edc   OPENIG-316 (CR-4524) Improve Heap interface usability ... Browse Dir » 
    Heap.get(String):Object gains type safety

    Heap.get(String, Class<T>):T

Heap.getRequiredObject(JsonValue, Class<T>) has a shorter name

    Heap.resolve(JsonValue, Class<T>):T

Heap.getObject(JsonValue, Class<T>) is replaced by a resolve
variant that supports optional dependencies:

    Heap.resolve(JsonValue, Class<T>, boolean):T

git-svn-id: https://svn.forgerock.org/openig/trunk@572 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     
  • a6e0cd60   OPENIG-311 OPENIG-316 (CR-4524) Support inlined declaration in the JSON configuration ... Browse Dir » 
    Inline object declarations are a mean to ease understanding of
Exchange processing.
They permit to describe anonymously, inner objects when a reference to
another heap object is required. That introduce, in the configuration
files, some hierarchical support, easing the user to mentally
represents his processing chain.

This is done in a fully backward compatible way, without requiring any
changes to existing object declarations (the one that requires other
objects through references or names).

The idea is to automatically extract inline declaration when the Heaplet
is calling the get***Object() methods: if the provided JsonValue is a
String, traditional object lookup is performed, but when the JsonValue
represents a JSONObject (a Map), we try to turn this into a normal
object declaration (just like what is done during heap initialisation).
If the given JsonValue does not describe a valid declaration, a
JsonValueException is thrown (again, just like the heap init process
is doing).

Notice that inline declarations do not require a 'name' attribute to
be specified (like anonymous Java classes), so we generate a unique name
based on the JsonPointer (represents the location of the node in
the JSON structure).

Notice that OPENIG-316 is partly resolved in this commit: HeapUtil methods
have only been moved into the Heap interface: no additional type safety,
no renaming.

git-svn-id: https://svn.forgerock.org/openig/trunk@571 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     

10 Sep, 2014

1 commit

  • 6e705541   OPENIG-278 CR-4474 Implement OpenIG Session using JWT and stores it in a cookie ... Browse Dir » 
    OpenIG used to provide a default Session implementation based on the
underlying Servlet container's HttpSession. This changeset intends to
gives to the user the ability to change the session persistence
strategy (in other words: changing the Session implementation).

This can be done at the global level (in the config.json, declaring a
SessionFactory object named 'Session') or on a per-route basis (with the
new 'session' attribute).
When an Exchanges comes into a route that declares a new session type, a new
session is build (no existing session items are propagated) and replace
the old session. When the exchange exits the route, the new session is closed
(notify the session that it's time to persist its content) and is replaced
by the old one. Really like a push/pop stack mechanism.

Notice that the 2 sessions are completely separated (cannot access the old
content from the new and vis-versa). First, that would defeat the purpose of
different session persistence modes (if items are propagated, where should
I persist them ?). Secondly, Session is not intended to share data between
handlers/filters: the Exchange is basically a request-scoped Map that is
designed for that purpose.

The JWT based session is a session implementation whose persistence is done
using an HTTP Cookie, the session's content being serialized as JSON (usable
types are constrained, see list below) and used as the payload of an
encrypted JSON Web Token (JWT).

The use of the JWT session has a few constraints:
 * HTTP Cookies are size-limited to 4K -> Small objects can be stored
 * Only JSON compatible types are supported:
    * null
    * Java primitive types + their boxed equivalent
    * Strings (and any CharSequence)
    * List and Map (of the supported types, recursively)
 * Same client performing concurrent HTTP invocations (so within the
   same HTTP session) that will modify their own session content will
   see inconsistencies in the session. This is due to the fact that the JWT
   session is not shared, each concurrent Thread has its own instance and
   can modify it at will. At the end of the processing, each Thread will
   serialize its own session's content regardless of other Threads.

git-svn-id: https://svn.forgerock.org/openig/trunk@565 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     

08 Sep, 2014

2 commits

05 Sep, 2014

4 commits

  • c0131ed2   Transform an array of references into objects from the heap ... Browse Dir » 
    This change eases references resolution when they are provided as a JSON
String array (like ["RefOne", "RefTwo"]).

git-svn-id: https://svn.forgerock.org/openig/trunk@559 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     
  • 8dec1539   Expose new security heap objects (KeyStore, KeyManager and TrustManager) ... Browse Dir » 
    KeyStore and related security objects are being more and more central
to OpenIG:
 * Used by HttpClient
 * Needed for encryption / decryption

This patch is a first step to provide a way to configure theses as
usual heap objects (through JSON).

KeyStore heap object is an abstraction around the keystore file (either JKS,
PKCS12 or other depending on the platform capabilities).
Specifying a password or not depends on he usage:
 * If KeyStore will be used to read private credentials (through a
   KeyManager for example), a password is required
 * If not (for a TrustManager), no password is required

KeyManager and TrustManager are using a KeyStore heap object and
are used within SSL mechanism (SSLContext) to provide a view on the KeyStore:
 * KeyManager when private credentials access is required.
 * TrustManager when only public information access is required

Issues: OPENIG-295
Reviews: CR-4441

git-svn-id: https://svn.forgerock.org/openig/trunk@557 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     
  • 8cb26e78   Provide the ability to evaluate JsonValue containing expressions 'in place'. ... Browse Dir » 
    The previously introduced evaluate() method was a step in the right
direction, allowing resolution of static (no references to ${exchange})
string JsonValue, returning a simple resolved String.
While it works great, it's a shame that we don't benefit anymore of the
JsonValue API, in particular the conversion methods (asURL(), ...).

This fix uses a JsonTransformer to actually resolve any String wrapped
JsonValue and creates a deep copy of the given JsonValue.

git-svn-id: https://svn.forgerock.org/openig/trunk@556 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     
  • 7a460545   Introduce HeapletFactory service support ... Browse Dir » 
    When a new heap object instance has to be created, we first have to
find a Heaplet instance that will create the required object from the
given JSON configuration.
Heaplet instances are linked to the single object they created: they
manage the heap object lifecycle with the start() and destroy() methods.
The old behaviour was using Heaplet instances directly loaded and
instantiated, meaning that a single Heaplet instance could be used to
create multiple heap objects, somehow mixing states, which was very
wrong.

Now, HeapletFactory (instead of direct Heaplet) are found using the
ServiceLoader discovery mechanism. They are responsible to create new
Heaplet instances, meaning that we cannot anymore share a Heaplet
instance for 2 or more heap objects.

Heaplet class does not extends Indexed<Class> anymore (was only used
because of the way Heaplets was looking for Heaplet instances).
NestedHeaplet is now deprecated since its only duty was to implement the
Indexed interface. All references to NestedHeaplet were replaced to
GenericHeaplet (its parent class) in our code base to avoid ugly
compiler warnings.

As per OPENIG-302 comments, the HeapletFactory will be a short-lived
concept since this additional layer introduction was the straw that broke
the camel's back :)

Issues: OPENIG-302
Reviews: CR-4457

git-svn-id: https://svn.forgerock.org/openig/trunk@555 dbb9e58e-28e6-4ce0-90e8-f11d9605b710
    guillaume.sauthier
     

26 Aug, 2014

1 commit

25 Aug, 2014

1 commit

21 Aug, 2014

1 commit

08 Aug, 2014

2 commits

04 Aug, 2014

1 commit

03 Aug, 2014

7 commits

31 Jul, 2014

2 commits

30 Jul, 2014

4 commits

29 Jul, 2014

4 commits

28 Jul, 2014

3 commits

25 Jul, 2014

2 commits

24 Jul, 2014

2 commits