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

[Galder Zamarreño]

On Jan 4, 2011, at 5:12 PM, Eduardo Martins wrote:

> Hi Galder, more comments inline:
> 
> On Tue, Jan 4, 2011 at 10:30 AM, Galder Zamarreño <galder at redhat.com> wrote:
>> See comments inline:
>> 
>> On Jan 3, 2011, at 3:44 PM, Eduardo Martins wrote:
>> 
>>> See inline.
>>> 
>>> On Mon, Jan 3, 2011 at 10:38 AM, Galder Zamarreño <galder at redhat.com> wrote:
>>>> See below my comments:
>>>> 
>>>> On Dec 30, 2010, at 7:05 AM, Eduardo Martins wrote:
>>>> 
>>> I actually don't know, it was not made due to an execution error. I
>>> saw there an Externalizer, I assume it's going to be "externalized",
>>> which means our DefaultConsistentHashExt that replaces it will also
>>> be.
>>> 
>>> Out of curiosity, here is the code, it forces colocation of certain data:
>>> 
>>> http://mobicents.googlecode.com/svn/branches/servers/cluster/2.x/infinispan/src/main/java/org/mobicents/cluster/infinispan/distribution/DefaultConsistentHashExt.java
>> 
>> That code is a bit strange. If you need to make sure that the hash from InfinispanClusterDataKey is derived from its key's hash depending on whether the key is DependentClusterDataKey or something else, why not just adjust  InfinispanClusterDataKey.hashCode() accordingly? i.e.
>> 
>> InfinispanClusterDataKey.hashCode() {
>>  if (getKey() instanceof DependentClusterDataKey) {
>>    return ((DependentClusterDataKey) clusterDataKey).dependsOn().hashCode();
>>  } else {
>>    return getKey().hashCode()
>>  }
>> }
>> 
>> I don't think you need a different consistent hash implementation.
>> 
> 
> hashCode() is not used only by the hash impl, it should be as unique
> as possible elsewhere.

So? I still don't get why your hashCode() impl should not represent the hashCode() of what's really relevant for InfinispanClusterDataKey...

>>> 
>>> So right now, there is no work being done to plug Infinispan in AS7?
>> 
>> I can't say 100% for sure, Paul Ferraro or Scott Marlow will be able to clarify this further, but AFAIK, this is not being done yet.
>> 
> 
> Would be good to have a roadmap for this, to adjust our work too.

Again, this is something you need to ping the AS guys about: Paul F, Scott M and if needed Jason too.

--
Galder Zamarreño
Sr. Software Engineer
Infinispan, JBoss Cache