Sorry about the late reply, Hotmail decided to put all replies
to this topic in the spam folder :-/
Holger Hoffstätte wrote:
> On 28.02.2014 10:28, Mike Wilson wrote:
> > Can you recommend any Remote Service distribution provider
> > implementations that offer better support for keeping "referential
> > integrity" within the data transferred to the remote server?
>
> I would love to hear what semantics of the consequently necessary
> distributed GC you'd expect in the face of failures, netsplits and
> stragglers. Once that is cleared up we can talk about why you
> think that
> distributed globally consistent object networks with (almost
> guaranteed)
> cycles are a good thing in reality to start with.
Where did I write that I want distributed globally consistent
object networks?! I didn't, and I don't.
I wrote "referential integrity *within* the transferred data",
meaning f ex that if I call a remote service with
MyClass a = ...
MyClass[] arr = new MyClass[]{a, a};
remoteService.call(arr);
with CXF I get this on the remote side:
[copyOfA, anotherCopyOfA]
while I want this:
[copyOfA, copyOfA]
Call semantics are completely by value, with no remote references
to deal with, just keeping cardinalities the same as in the original
data. It's up to how the data is serialized. Most serializers
inspired by java.lang.Serializable support this (f ex XStream) but
not the ones coming primarily from the XML world (f ex CXF).
> There's a reason (many actually) why nobody with half a brain
> (get it?!) uses RMI any more.
When in doubt, assume the OP has only half a brain? Great.
> > Bonus question: What's a good setup for load balancing
> Remote Service
> > calls between multiple remote servers?
>
> "For stateless or stateful services?" he asked, tongue in cheek.
Stateless, of course.
Mike
_______________________________________________
OSGi Developer Mail List
[email protected]
https://mail.osgi.org/mailman/listinfo/osgi-dev
