- From: Cutler, Roger (RogerCutler) <RogerCutler@ChevronTexaco.com>
- Date: Sat, 5 Oct 2002 10:28:19 -0700
- To: www-ws-arch@w3.org
- Message-ID: <7FCB5A9F010AAE419A79A54B44F3718E2EAF43@bocnte2k3.boc.chevrontexaco.net>
I have a question that I think is about web services architecture. Let me first state what I think the situation is and then ask the question. If this is a relevant question I suspect that the folks on this list may be able to recast it to be more meaningful. *** The Situation (I think?) *** I think that WSDL does not explicitly expose objects on the web. One could imagine a web service protocol (which might be part of a web services architecture??) that does this, but WSDL is not it. There is nothing in WSDL, for example, that explicitly or even implicitly supports "Dog and Cat interfaces implement the Animal interface". Moreover, WSDL does not explicitly support recognizing the following three calls and sending them to different methods appropriately: Foo(int) Foo(int, int) Foo(float) Now I know someone who thinks he's seen the first two implemented in .Net and I find this very believable. I'd even think it conceivable that the third might work, although doing so would be quite tricky. But making the Foo's work depends on the implementation, not design features in WSDL, so interoperability depends on implementation -- which experience shows has been quite variable. I think it's quite possible that however .Net does the first two might also work on other platforms, since most implementations are on platforms that support that kind of overloading, but if you wanted to depend on it you'd have to find out specifically, and that kind of question is not real easy to get answered. *** The Question *** If the above observations are more or less correct, does it make any sense to treat the possibility of exposing objects via web services in the web services architecture?
Received on Saturday, 5 October 2002 13:28:42 UTC