Friday, May 31, 2002

Matt Griffith wonders if I know...

Yep. Thus, WS-Inspection. :-)

Wednesday, May 08, 2002

Paul Prescod's rules for Extreme Web services. I agree with #1.

Edited my previous statements on section 5. Thinking about it some more, I do wish we had every reasonable feature that section 5 and the SOAP spec allows (multidim is at the top). However, real-world engineering has limited that goal in V1. V1.1 is a little better. And V2 will even be better than that. The SOAP Toolkit has been a good model for that (especially since they've been able to ship so often.)

As an aside, I use to feel that subsets were a bad idea. At this point, I don't think interop can happen without them. I hope that the WS-I basic profile gives us a reasonable one involving WSDL, SOAP, XSD, etc...

Simon says: "Outside of section 5, header faulting is not handled correctly [it will generate a mustUnderstand fault, even if the header is not targeted at the ultimate destination]".

Actually, we leave it to the app developer to decide because actor resolution is such a sticky issue. They can set the DidUnderstand property to true, and then we won't fault. Otherwise, we err on the side of caution and fault.

BTW: aren't sparse arrays the same as partial arrays?

Tuesday, May 07, 2002

Simon says: "my experience the MSTK is significantly more compliant with the specs than ASP.NET web services."

My experience (and the interop results from SOAPBuilders) is that both products have very few interop bugs. And where there are reported interop failures with ASP.NET Web services, it's because we didn't support a certain section 5 features (specifically: HexBin, sparse arrays, and multidimensional arrays). If having those features equals "signifigantly more compliant", I disagree.

We default to literal, because we feel that it is much better than encoding.

Sunday, May 05, 2002

Simon's been reading some great SF.

Little known fact: I live about 1 mile from Eric Nylund. I use to work with his wife Syne at Microsoft.

Thursday, May 02, 2002

Simon catches me in a small gaffe. Indeed, the QName of the root element is what really matters.

However, I don't think I'm understood. I don't mean, "how do I tell that a piece of XML is a SOAP message", but given any SOAP message, how I tell the "type" of that SOAP message? In otherwords, what


  • message in which
  • operation of what
  • portType in which
  • WSDL?

Wednesday, May 01, 2002

New dangerous question of the day: What is the type of a SOAP message?

Given any XML document, you can look at it's default namespace and tell it's type. For instance, a WSDL 1.1 document has the WSDL namespace as it's default namespace, and thus you can identify that XML document as having the type "WSDL". What's the type identifier for a SOAP message?