I'm considering api that may be required for improved OSGi support.
In order for an OSGi environment to deserialize a proxy, it needs to
first install a bundle for the proxy and resolve any dependencies. For
OSGi a ProxyPreparer must first locally marshall (create a
MarshalledInstance) a java.lang.reflect.Proxy (implementing
ServiceProxyAccessor, ServiceCodebaseAccessor,
ServiceAttributesAccessor) instance (returned by
SafeServiceRegistrar.lookup) and unmarshall it, passing in the bundle
ClassLoader as the default ClassLoader. This ensures the ServiceProxy's
Bundle ClassLoader becomes the default ClassLoader for the underlying
JERI endopoint. Only at this time will a call
ServiceProxyAccessor.getServiceProxy() result in a correctly
unmarshalled proxy. If this step isn't performed the default
ClassLoader for the JERI Endpoint will be the SafeServiceRegistrar's
proxy ClassLoader, and a ClassNotFoundException will be thrown when
calling ServiceProxyAccessor.getServiceProxy().
Given there's some complexity in the above, it would be prudent to
implement this in say a convenience class, perhaps called
OSGiProxyPreparer, so developers don't have to (boilerplate).
But we still need something from the underlying modular framework, to
install a Bundle for the service proxy and to ensure OSGiProxyPreparer
recieves a ClassLoader, while avoiding a dependency on OSGi. The
OSGiProxyPreparer could accept a ProxyClassLoaderProvisioner (see below)
as a constructor argument? Keep in mind the ProxyPreparer is a
configuration concern.
The discovery infrastructure (LookupLocator, LookupLocatorDiscovery and
LookupDiscover classes) also needs a way to receive a ClassLoader to
deserialize lookup service proxy's. The codebase URL can be provided in
a multicast response, the same interface would need to be used as in
ProxyPreparation.
Please provide feedback, thoughts or suggestions.
/*
* Copyright 2017 The Apache Software Foundation.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package net.jini.loader;
import java.io.IOException;
/**
* Allows client code to implement provisioning of a ClassLoader
for a URI path,
* where the use of codebase annotations is unsuitable.
*
* The first URI in the path, must be the proxy URI, any
additionally appended
* URI are dependants, it's possible these may not be loaded
directly by the returned
* ClassLoader, but their classes should still be visible and
resolvable from it.
* Only the proxy URI classes are guaranteed to be loaded by a
returned
* ClassLoader.
* Dependant URI are not guaranteed to be loaded if suitable
versions are
* already.
*
* Some systems, notably OSGi, manage ClassLoader visibility
differently than
* Java's typical hierarchical ClassLoader relationships,
implementors of this
* interface may implement this to provision codebases into a
ClassLoader,
* prior to deserializing a proxy into the resolved ClassLoader.
*
* A proxy may be an instance of {@link java.lang.reflect.Proxy}
and have
* no associated codebase, in this case, a new ClassLoader should
be returned
* into which a dynamic proxy class can be provisioned.
*
* The implementing class must have {@link java.lang.RuntimePermission}
* "getClassLoader" and "createClassLoader". The implementing
class must
* ensure that the caller has {@link java.lang.RuntimePermission}
* "createClassLoader" as well.
*
*/
public interface ProxyClassLoaderProvisioner {
/**
* Create a new ClassLoader, given a space separated list of URL's.
*
* The first URL must contain the proxy class.
*
* The parent ClassLoader may be ignored, depending on the
underlying
* infrastructure.
*
* If uriPath is null, a ClassLoader is still created to
accommodate
* any dynamically created classes, with the parent ClassLoader
as the
* parent.
*
* @param uriPath the codebase URL path as a space-separated list
* of URLs, or <code>null</code>
* @param parent the default parent ClassLoader.
* @return
* @throws IOException
* @throws SecurityException if the caller doesn't have {@link
java.lang.RuntimePermission} "createClassLoader".
*/
public ClassLoader createClassLoader(String uriPath,
ClassLoader parent) throws IOException;
}
Regards,
Peter.
On 27/09/2017 6:59 PM, Peter wrote:
Some updates on thoughts about OSGi:
1. In JGDMS, SafeServiceRegistrar (extends ServiceRegistrar),
ServiceDiscoveryManager and ProxyPreparer allow provisioning of
OSGi bundles for Jini services.
2. SafeServiceRegistrar lookup results contain only instances of
java.lang.reflect.Proxy (implementing ServiceProxyAccessor,
ServiceCodebaseAccessor, ServiceAttributesAccessor) which a user
remarshalls and unmarshalls into their OSGi bundle provisioned
ClassLoader, prior to retrieving the actual service proxy using
ServiceProxyAccessor.
3. As a result different service principals using identical proxy
codebases, needn't share a ClassLoader, addressing the trust
domain issue previously alluded to.
4. There's no current mechanism to allow provisioning of a bundle for
a Registrar.
5. Existing discovery providers accept ClassLoader arguments for
unmarshalling Registrar's.
6. Existing Multicast responses allow for additional information to
be appended; a codebase resource for example.
7. LookupLocator, LookupDiscovery and LookupLocatorDiscovery classes
don't utilise discovery providers ClassLoader arguments.
8. Need to allow bundles to be provisioned for lookup services after
multicast discovery, by exposing discovery provider ClassLoader
arguments and allowing client to manage provisioning of bundle
into a ClassLoader, then passing that in during unicast discovery.
9. Don't break backward compatiblity.
Cheers,
Peter.
On 16/11/2016 4:18 PM, Dawid Loubser wrote:
+1 for OSGi providing the best solution to the class resolution problem,
though I think some work will have to be done around trust, as you say.
On 16/11/2016 02:23, Peter wrote:
The conventional alternatives will remain; the existing ClassLoader
isolation and the complexities surrounding multiple copies of the
same or different versions of the same classes interacting within
the same jvm. Maven will present a new alternative of maximum
sharing, where different service principals will share the same
identity.
Clearly, the simplest solution is to avoid code download and only
use reflection proxy's
An inter process call isn't remote, but there is a question of how a
reflection proxy should behave when a subprocess is terminated.
UndeclaredThrowableException seems appropriate.
It would plug in via the existing ClassLoading RMIClassLoader
provider mechanism, it would be a client concern, transparent to the
service or server.
The existing behaviour would remain default.
So there can be multiple class resolution options:
1. Existing PrefferedClassProvider.
2. Maven class resolution, where maximum class sharing exists. This
may be preferable in situations where there is one domain of trust,
eg within one corporation or company. Max performance.
3. Process Isolation. Interoperation between trusted entities,
where code version incompatibilities may exist, because of separate
development teams and administrators. Each domain of trust has it's
own process domain. Max compatibility, but slower.
4. OSGi.
There may be occassions where simpler (because developers don't need
to understand ClassLoaders), slow, compatible and reliable wins over
fast and complex or broken.
A subprocess may host numerous proxy's and codebases from one
principal trust domain (even a later version of River could be
provisioned using Maven). A subprocess would exist for each trust
domain. So if there are two companies, code from each remains
isolated and communicates only using common api. No unintended code
versioning conflicts.
This choice would not prevent or exclude other methods of
communication, the service, even if isolated within it's own process
will still communicate remotely over the network using JERI, JSON
etc. This is orthogonal to and independant of remote communication
protocols.
OSGi would of course be an alternative option, if one wished to
execute incompatible versions of libraries etc within one process,
but different trust domains will have a shared identity, again this
may not matter depending on the use case.
Cheers,
Peter.
ESent from my Samsung device.
