HTTP Binding thoughts

I see that there are a number of issues related to HTTP binding.  Issues 64, 55, 54, 47, others? really seem to fit into a consistent theme of "how much of HTTP can WSDL describe".  Now clearly we have to do some work here, because SOAP 1.2 has the web method feature.  But I think there is also a consistent undertone of how much of just HTTP can WSDL describe.   

I argue that a good test case is whether WSDL can describe Atom's use of HTTP.  

I think I'm in favour of at least exploring what a somewhat full features HTTP binding would look like before deciding whether or not to limit the scope.  For example, I posted a strawman for issue 64 that provided a description of HTTP operations.  I have the feeling that this will help with meeting the s12 webmethod feature and be more broadly usable.

I'm a bit concerned that closing issue 64 and saying "forget about it" would leave us disadvantaged when looking at the rest of the issues.  And if we follow a similar pattern of closing early with out of scope on 55, 54, we may impede our ability to create a design that meets our requirements.  I'm also somewhat concerned that we might go down the "death of a thousand cuts" where we continually argue about scoping of various things, and then end up with something that allegedly meets our requirements but everybody is pretty unhappy with.  Maybe something like what happened with the grid request for property styles.  

My request is that on issues of HTTP binding, we do not close them early, but rather we look at them in the context of concrete design proposals.  I'm thinking that a few turns of the design crank will let us know whether the related issues can be satisfied or not.  I really do have the feeling that there is a design lurking that will meet our requirements and potentially more, and people will be happy with it.  

Cheers,
Dave

Received on Wednesday, 7 April 2004 17:19:43 UTC