Re: [infinispan-dev] [Pull Request] Modular Classloading Compatibility
BTW having thought about it more, this seems to be the only option that actually works at
the API level, even if it will require some re-eng in Infinispan core.
On 16 May 2011, at 19:29, Pete Muir wrote:
On 16 May 2011, at 19:24, Sanne Grinovero wrote:
> 2011/5/16 Pete Muir <pmuir(a)redhat.com>:
>> This is the "hibernate session style contract" that Jason is talking
about. As the CM can be shared (e.g. in JNDI), then the same Cache object can be returned
in multiple applications in the app server, meaning you can't simply associate the CL
with a Cache object when it's created.
>
> A CM could be JNDI registered, but a Cache is not registered. assuming
> same cacheManager, two different applications could do:
>
> cacheA = cm.getCache( "hibernateCache", appAClassLoader );
> cacheB = cm.getCache( "hibernateCache", appBClassLoader );
>
> cacheA != cacheB
> still they will delegate toe the same "hibernateCache", but being
> different only in which ClassLoader they set in their invocation
> context, which will be read by the Unmarshaller.
AIUI you just said what I said in different language?
_______________________________________________
infinispan-dev mailing list
infinispan-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev