[infinispan-dev] Feedback from Mobicents Cluster Framework on top of Infinispan 5.0 Alpha1

[Manik Surtani]

Thanks for the feedback, Eduardo.  I'll address each one separately, although Galder's done an excellent job already.  :)

On 4 Jan 2011, at 16:12, Eduardo Martins wrote:

>> 
>> Oh right, I see what you mean. You're saying that getCacheManager() in Cache returns a CacheContainer rather than an EmbeddedCacheManager. There's a reason for this:
>> 
>> CacheContainer is the common demoninator for operations that are accessible by remote and local caches. EmbeddedCacheManager extends CacheContainer to add operations that only make sense on a local basis, whereas RemoteCacheManager does the same for remotely accessed Infinispan instances. For the moment, you cannot add listeners to remote cache/cachemanager instances, hence why those methods are only available in EmbeddedCacheManager.
>> 
>> It might make sense to have an EmbeddedCache interface that ties up this up via covariant returns, i.e.:
>> 
>> EmbeddedCacheManager getCacheManager();
>> 
> 
> Agree.

This was changed in 4.1.0 when we introduced the RemoteCacheManager (Hot Rod client).  But yeah I appreciate your concerns re: not being able to add a cache manager listener when all you have is a reference to a cache - however I would have expected that you'd also have a reference to an EmbeddedCacheManager (since you created the cache!) so you could add listeners there.

I don't think a covariant return type will work.  If Cache.getCacheManager() returns EmbeddedCacheManager, then RemoteCache.getCacheManager() cannot return RemoteCacheManager since it isn't a sub-interface of EmbeddedCacheManager.

TBH, RemoteCache.getCacheManager() is a no-op anyway - so I don't see why Cache.getCacheManager() doesn't just return EmbeddedCacheManager.

Galder, did you create a JIRA for this?

Cheers
Manik
--
Manik Surtani
manik at jboss.org
twitter.com/maniksurtani

Lead, Infinispan
http://www.infinispan.org