[XML-SIG] RE: SOAP for Python

Uche Ogbuji uche.ogbuji@fourthought.com
Wed, 02 May 2001 22:40:28 -0600

> > Until then, I rely on the fact that section 5 is explicitly optional.  There
> > is no requirement for a SOAP implementation to use the SOAP serialization.
> Technically right, but it would be *very* surprising and upsetting to
> folks who naively used the 4SS implementation to talk to other web
> services.  It might even cause them to spit venom at you.

After I sent my last message another thought struck me.  You use the term "Web 
services" above.  Probably I have to understand what you mean by that before I 
understand why you think it would be surprising and upsetting to have SOAP 
systems that don't implement section 5.

The only reason everyone would want to "just stick to section 5" is for 
"transparent" API-type calls.  RPC all over again.  Basically CORBA with 
SOAP/HTTP over the wire rather than CDR/IIOP.

But what on earth is the use of such a thing?  Why not just use CORBA or DCOM 
or RMI, all of which are vastly more efficient than SOAP and can claim more 
pedigree and interop?

The answer is simple: because such tightly-coupled systems do not survive the 
boundary from one business technology and process to another.  Crossing such a 
boundary requires loosely-coupled systems, and that is the only reason there 
is any relevance to the buzzword "Web services".

Successful Web services will be message-oriented, loosely coupled systems with 
a great deal of flexibility that is handled through metadata management.  
Whether you're in the ebXML camp or the UDDI camp, you had better be taking 
those tModels, WSDL bindings and CPPs seriously, because if you just blindly 
write code that assumes that, say everyone uses SOAP serialization, you will 
be doing commerce with only a fraction of your brave new market.

This is why it was utter silliness for section 5 not to have been broken out 
of SOAP transport into a separate spec.  It encourages people to wrongly 
assume that SOAP implies section 5, and thereby condemn themselves to 
reinventing the RPC wheel all over again.

And I'll note that I'm not alone in this sentiment.  In past SOAP debates on 
XML-DEV, no lesser figures than Tim Bray and David Megginson have expressed 
similar annoyance at the conflation of transport and content model that mars 

So do I think it's realistic that section 5 will be put in its place before XP 
emerges?  Absolutely.  And in the unlikely event that this doesn't happen, Web 
services will pretty much drown in its own unfulfilled promises.

Uche Ogbuji                               Principal Consultant
uche.ogbuji@fourthought.com               +1 303 583 9900 x 101
Fourthought, Inc.                         http://Fourthought.com 
4735 East Walnut St, Ste. C, Boulder, CO 80301-2537, USA
Software-engineering, knowledge-management, XML, CORBA, Linux, Python