[Important Note: - AMQP is NOT and interoperability protocol in the same way that NFS, SMTP and HTTP are not interoperability protocols. If AMQP were an interop protocol it would have a very wide tolerance for semantic variation, and a lowest common denominator sweet spot; which is definitely not the case. If you want interop, look at WS-I.]
AMQP compliance testing is an important, and expensive area. Several approaches have been discussed at the AMQP WG, and the WG is definitely the place that should own a formal compliance test (and it should be freely accessible). There are discussions about who can contribute what to that over at the WG, and it wouldn't be surprising if tests from Qpid and other projects were volunteered as parts of the compliance test. So, the immediate reality of this is that "Connectathon" style interop is going to have to be the first port of call. The JMS Qpid client talks to the OpenAMQ C server today, which is a good start. When the Qpid C++ and Java servers are going there are lots of interop tests we should be doing; and if/when ActiveMQ implements the protocol semantics we have another dimension of testing. It is important to the success of AMQP open middleware that people take interop seriously -- the market for AMQP depends on this and people who want to make $$ off AMQP depend on a growing market. That's the theory, just think NFS/HTTP/SMTP for a model. When implementations stabilise at a suitable level of functionality, I'd imagine I'll resource building some AMQP WG protocol compliance tests -- since its very much in everyones interests to have them. Finally, I'm hoping AMQP may become the common transport protocol for many of the financial services initiatives going on - like MiFID, SEPA, or the ISO 20022 work. AMQP is only really valid for that kind of work if that standard is supported by several products which can verify their compliance -- for those of you who have heard of FIX or SWIFT you'll know they have compliance test suites. If anyone wants to volunteer tests at this stage I'm sure they would be well received, but I think we have to rev the AMQP spec itself through another draft before we consider it. Cheers John <with my AMQP WG hat on :-) > On 20/10/06, Carl Trieloff <[EMAIL PROTECTED]> wrote:
I would say that as AMQP publishes test, Qpid run them or add implementations to written definitions into the pyhton test harness. However this project is not about creating test suites, but about creating a complete implementation of AMQP, which will by definition include tests. Hiram Chirino wrote: > If AMQP is supposed to be a message provider interropability protocol, > shouldn't qpid be focused on implementing Broker and Client compliance > tests > suites? >
