60 likes | 170 Views
GGF 16 InteropFest BoF. Glenn Wasson University of Virginia. InteropFests In Which We Participated. WSRF / WSN InteropFests 2003 at IBM, Raleigh, NC 2005? via the web Based around scenario documents Running example that involves specs messages. InteropFests In Which We Participated.
E N D
GGF 16 InteropFest BoF Glenn Wasson University of Virginia
InteropFests In Which We Participated • WSRF / WSN InteropFests • 2003 at IBM, Raleigh, NC • 2005? via the web • Based around scenario documents • Running example that involves specs messages
InteropFests In Which We Participated • First was a kick-off to the OASIS WGs • “Show of support” in the formation of the WGs • Roughly a dozen independent implementations (in various states) • Equal mix of industry and academia/government • Second was pre-public comment testing • Last check before “release” • 6 (ish) implementations • 75% / 25% industry to academia/government
First InteropFest • “Developers in a room” variety • Scenario document described specific tests • Matrix of check boxes • What was interesting? • Understanding of time (.NET could parse lots of serializations of time, but only generate one) • Have to push this back in the spec (i.e. you must accept time in the following formats) • Small changes in messages can require large changes in infrastructure
Second InteropFest • Remote “Publish your URLs” variety • Again, had scenario document • More disconnected / Less frenetic coding • What was interesting? • Harder to test clients • Theory → everyone tests their own clients against your service • Practice → it’s easier to run everyone else’s clients against your service
Lessons? • Dev-in-room more productive per unit time • Typical F2F vs. telecon trade-offs • Spec-wise legal WSDL / XSD ≠ Good • Without tooling support, forget it • Interop tests can find these issues • Easy to get caught in “client/service on same platform” trap • Spec changes can come out of interop tests! • This is what we want… but spec authors can resist…