Uche Ogbuji has doubts about the value of SOAP:
Jonathan Bartlett has a spot-on rant on why CORBA is still superior to SOAP. I think a lot of people know this, but fear to say it. Duncan Grisby, the lead developerof the superlative omniORB came to the tenth International Python Conference (IPC10) and was on a Web services panel. I think he came expecting to have to defend CORBA's honor and was surprised when just about all the panelists agreed with him that CORBA is better than SOAP in practice.
He wasn't done there; he posted later on SOAP vs. CORBA:
This is a more in-depth comparison of SOAP to CORBA than the rant I blogged earlier. It is quite sharply biased against SOAP (I've never been much swayed by "eeew XML is so verbose" arguments). But I think many of its points are fundamentally sound. Mike Olson also ran some SOAP/XML-RPC/CORBA performance tests on Python, with remarkable results.
Those benchmarks came out with SOAP being dramatically slower than CORBA. Not a big surprise; sending reams of uncompressed text over the wire - as opposed to a binary protocol - isn't going to be fast. But he wasn't done yet - read here as well:
Rem acu tetugit Paul Prescod. IOW, He nailed the point, as he so often does. I think the idea of merging header and body into a single document is the single biggest flaw in SOAP. Yes, SOAP section 5 (the RPC datatyping section) was probably the largest overall mistake in SOAP's evoution (as even heavyweight SOAP boosters have started admitting now), but it is far less fundamenal than SOAP's monolithic design.
We (Cincom Smalltalk) support SOAP well; it's got a bandwagon of support, and there's no way we can just tell developers to use CORBA instead. Still, Uche makes some very good points here - well worth considering if your technical options have not been closed off ahead of time....