On Wed, Mar 18, 2009 at 3:50 AM, Luciano Resende <[email protected]> wrote: > On Mon, Mar 16, 2009 at 1:55 PM, Yang Lei <[email protected]> wrote: >> Hello , >> >> I read the proposal for HTTP binding specification : >> >> http://lists.oasis-open.org/archives/sca-bindings/200810/msg00078.html >> >> I wonder if we can support part of the specification so that we can have a >> generic binding based on HTTP protocol , use the wireFormat and >> operationSelector definition to support a specific mapping between >> messageBody and SCA services... >> >> The part of the schema interests me is: >> >> <binding.http uri=”xs:anyURI”? ...> >> <wireFormat/>? >> <operationSelector/>? >> <response> >> <wireFormat>? >> </response>? >> </binding.http> >> >> One sample of using this binding will be converting the JSON-RPC binding to >> be an HTTP binding with the wireFormat of JSON-PRC: >> >> <component name="Catalog"> >> <implementation.java class="services.FruitsCatalogImpl”/> >> <service name="Catalog"> >> <t:binding.http uri="http://localhost:8080/Catalog";> >> <wireFormat.jsonrpc/> >> </t:binding.http> >> </service> >> </component> >> >> <reference name="catalog" target="Catalog“> >> <t:binding.http> >> <wireFormat.jsonrpc/> >> </t:binding.http> >> </reference> >> >> I noticed there are some discussion on the list last year on this: >> http://mail-archives.apache.org/mod_mbox/tuscany-dev/200810.mbox/%[email protected]%3e >> >> Appreciate your comments. >> >> Regards, >> >> Yang Lei >> >> > > I think we could definitely experiment on that direction as it seems > the direction being proposed to OASIS. > > One thing to have in mind is that this might impact the functionality > existent in the current binding.http, so it might be good to start > this effort as a fork/copy of the binding, and decide which one would > be named binding.http. > > What others think ? >
+1 on implementing the draft http binding spec. Implementing it will help us give feed back on the spec to OASIS to make the final spec good, and it seems like useful function to me. Along with a jsonrpc wireformats it would be interesting to also have others like jsonp, pox, raw, etc. Its worth noting that in the user survey we did a while back of all the bindings we have the Tuscany http binding was the second most popular with only the WS binding being more used. Hard to tell what people are doing with it, i've wondered if that was a survey mistake, or maybe they use it to have have an SCA component with access to the http request/response which has been asked for a few times on the user list. Co-existence and compatibility with the existing Tuscany http binding is an interesting question. They are in separate namespaces so we could in theory have them both, though that seems bound to result in errors when the wrong namespace or modules are included. Maybe in 2.x we could say we only support the OASIS http binding spec, though that doesn't quite fit in with whats being said about backward compatibility. ...ant
