Re: [ws-tx] Issue 007 - WS-C: Make Register/RegisterResponse retriable

[Kevin Conner <Kevin.Conner@jboss.com>]

Alastair Green wrote:
> A crashed participant should only reregister if its state is "intend to 
> register, but have not received RegisterResponse". If it has registered 
> then it should just come up and wait, or replay its state-appropriate 
> message.

That was the intention as per your scenario, of course it could be 
extended to include reregistration should it come up on a different 
endpoint :-)

> If the participant has a generated a GUID, and uses it to replay its 
> registration (wherever the identity is encoded) then it will never "clash".

Provided the algorithm to generate the GUID is perfect and doesn't 
generate clashes.  You obviously have much more faith in these 
algorithmg than I :-)

> We cannot tell the difference (and have no need to tell the difference) 
> between an inadvertent (transport) double delivery, an advertent retry 
> without crash, where the participant gets worried it never got the 
> message through, and a retry after crash recovery.

But we do need to tell the difference as the request may have come from 
a different participant that doesn't have the benefit of perfect GUID 
creation :-)

> They are all the same thing from the protocol's standpoint. Too much 
> information, otherwise.

Sorry Alastair but I have to disagree :-)

	Kev