I
agree.. same socket bi-directionality is exotic and does not have to be in the
public interface.
Currently, it's not very nice how the proxy going back client gets
setup. I won't go into details about the current way I do it (it
sucks).
Here's the route I
hope I can change too soon:
1) client sends a proxy object down an invocation to
the server..
2) server invoker sets either (a) an
Invocation attribute or (b) and ThreadLocal to identify the socket the
invocation came over.
3) When the proxy object is de-serialized, he looks
up the object in (2) to setup the callback channel to go through the original
socket.
But for this to work in the (a) case the
proxy object would need to be able to access the Invocation object during
deserization.
And for it to work in the (b) case the proxy object
has to be deserialized while in the context of the thread the invocation came
in.
I don't think I'll be able to do this the way
invocations are done right now. You think this is a bad route?
Regards,
Hiram
-----Original Message-----
From:
[EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]On Behalf Of Bill
Burke
Sent: Sunday, November 17, 2002 11:16 PM
To:
[EMAIL PROTECTED]
Subject: RE: [JBoss-dev] new
PooledInvoker: speeds up invocations
I
agree Scott (no public interface for bi-directionality). It will be
tricky to implement the bi-directional behavior if Invokers don't have a
bi-directional public interface. I wonder how you abstract this out
now Hiram.
One way to do this would be to use the trick I used with EJBs and
multiple-invokers.
1.
Have a proxy factory.
2.
Pass this information with the client request.
3.
Look up the proxy factory whenever you have to create proxies via the tag in
the invocation.
Bill
It should be possible to route invocations
over the incoming channel, but it cannot be
a requirement. This does not imply the
invocation interfaces are bi-directional however.
There simply needs to be the ability to
compose the invokers such that there is a shared
communications channel.
xxxxxxxxxxxxxxxxxxxxxxxx
Scott Stark
Chief Technology
Officer
JBoss Group, LLC
xxxxxxxxxxxxxxxxxxxxxxxx
----- Original Message -----
Sent: Friday, November 15, 2002
7:11 PM
Subject: Re: [JBoss-dev] new
PooledInvoker: speeds up invocations
Hiram Chirino wrote:
Hiram Chirino wrote:
> Anyways. JMS need bi-directional invocations (BADLY). Should this
> become a requirement for the other invokers??
I completely disagree. There is no reason server to client
communication has to go over the back channel of a client to server
I might have said this before, but there is one reason it's a nice feature:
This allows callback to clients that are sitting behind a firewall.
Let
me agree with Hiram. I would add it is essential to use the client
connection if the client network is using NAT (network address
translation). With the advent of new security measures, many of our
clients are adding firewalls and NAT. The worst case is when the server
has a public IP and the local LAN clients are on the otherside of a NAT
router.
--Victor
