[
Lists Home |
Date Index |
Thread Index
]
On Mon, Apr 22, 2002 at 03:20:45PM -0700, Dave Winer wrote:
> > To me, the matter of hours suggests we're looking at a trivial test
> > case, which doesn't prove anything either way.
>
> Hmmm. Consider this possibility instead. The SOAP toolkit they used has been
> patiently tested against many popular SOAP implementations, and that work
> wasn't for naught. When it came time to test the API, it worked.
So what. All that demonstrates is that it is possible to re-invent
XML-RPC with new tools. It also helps that the Google SOAP API is
playing in a very well-worn portion of SOAP that's easy to test
and interop.
I shudder to think what will happen when someone tries to publish
a SOAP API with a more complex schema. We won't get interoperability
by waving the magic WSDL pixie dust, and we'd be significantly
better off with some kind of task-specific XML document structure
with REST.
Amazon released an HTTP+XML API last week, and like the Google
interface, it's a very simple operation. We're not trying to
achieve interop with N x N peers; we're all trying to get *one*
service to understand us, and understand what that *one* service
is returning. It's not Google's (or Amazon's) responsibility to
test N different client implementations anyway; it's their
responsibility to publish a standard interface and adhere to it.
The "SOAP brings interop for free" argument is simply a straw man.
Especially when it necessarily prevents interop with a whole
class of tools (XSLT processors for one).
Z.
|
