Related link: ftp://www6.software.ibm.com/software/developer/library/ws-reliablemessaging20050…
WS-ReliableMessaging was republished on Tuesday of this week. The most substantial change was in the legal language, which grants royalty-free rights to implement. The most important technical difference is the relaxing of a restriction that used to require that a recipient keep track of message sequences beyond the receipt of a TerminateSequence message. The recipient is now allowed to disregard any straggling messages that arrive after a sequence has been terminated. For more information -
http://www-128.ibm.com/developerworks/webservices/library/ws-rmreload/
ftp://www6.software.ibm.com/software/developer/library/ws-reliablemessaging200502.pdf
Dave
Sequence protocol forces recipient complexity
I've recently been reading the WS-RM specs and they seem clear and well thought out.
The only thing that surprises me is that only one model of sequence resending seems to be supported.
The standard example message exchange (reproduced in most articles) shows how msg 2 in a 3 msg sequence is dropped and how this is then resent.
For this to be useful to the recipient it would have to implement a temporary store of incoming messages which it can use to fill in the gaps following resends.
In other protocols the recipient has the option of telling the sender that a set of messages were missing and that it would like all messages starting at the first missing message resent. In this way the recipient can be implemented without a temporary store. One example of a protocol like this is the FIX session protocol.
Do you know if these sort of models were considered?