The current code approach is not ideal.

It's about handling the deserialization of service proxies, in the
context of servlet container clustering.

When serialized, they store a token identifying the service id.

When deserialized, they need to find a Registry and replace the token
with the service proxy in the new Registry.

The servlet container is responsible for the deserialization, which
limits our options.  At startup a Registry places an instance into the
static field, since a static field is all that will be available to a
de-serializing service token.

Ideally, the servlet containers would post some kind of notification
before deseriallizing data from another server in the cluster; we
could then do something using a ThreadLocal, rather than a static.

Perhaps we need to do some research to see how to get access to that
kind of callback information, presuming it exists.

On Thu, Mar 25, 2010 at 12:48 PM, Igor Drobiazko
<[email protected]> wrote:
> Hello,
>
> before fixing TAP5-1078 [1] I would like to ask if there is any reason why
> the registry is stored in a static field of SerializationSupport. As
> described in [1] this causes duelling registries inside a JVM. I suggest to
> remove SerializationSupport and move the creation of ServiceProxyToken to
> InternalRegistry. Any objections? Am I missing something?
>
> [1] https://issues.apache.org/jira/browse/TAP5-1078
>
> --
> Best regards,
>
> Igor Drobiazko
> http://tapestry5.de/blog
>



-- 
Howard M. Lewis Ship

Creator of Apache Tapestry

The source for Tapestry training, mentoring and support. Contact me to
learn how I can get you up and productive in Tapestry fast!

(971) 678-5210
http://howardlewisship.com

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to