This problem doesn't occur on a method that doesn't return an array of
"complexTypes".  I have several other methods that return different types of
complexTypes, and those methods also work in either scenario.

For example this complexType (user):

<xs:complexType name="user">
    <xs:sequence>
        <xs:element name="userid" type="xs:int" />
        <xs:element name="firstname" type="xs:string" />
        <xs:element name="middlename" type="xs:string" />
        <xs:element name="lastname" type="xs:string" />
        <xs:element name="suffix" type="xs:string" />
    </xs:sequence>
</xs:complexType>

Is made into a class by WSDL, like this:

Public Class user
   
   Public userid As Integer
   Public firstname As String
   Public middlename As String
   ublic lastname As String
   Public suffix As String
   
End Class

The resulting method for returning an array of these was listed in my first
post.  I am trying to use the "contract first" method for design of the web
service.  I created an XSD with all my types and messages (one incoming and
one outgoing) for each operation.  I them created a WSDL using the XSD.  The
difference was that I ran WSDL.exe on a real WSDL, and "Add Web Reference"
ran it on a "asmx?WSDL" reference.  I generated the "server stubs" using
WSDL.exe i.e.

wsdl.exe /server /out:Server.vb /l:VB AppSec.WSDL AppSec.XSD

Then I placed this file in an "Empty Web Project", added a Web Service, made
the new Web Service class inherit the "Server" stub class, override the
methods, and that was it.

Enough rambling for me.  If I do all these steps correctly the client should
turn out the same each way.  What I am saying for these scenarios

- A server stub is generated from a WSDL (real one), and the client is also
generated from a WSDL (real one).
- A server stub is generated from a WSDL (real one), and the client is
generated via "Add Web Reference".

The client should be the same in each scenario, correct?

Well that question makes me wonder if the WSDL file is different from the
ASMX?WSDL

I did a quick comparison of these two:

- In my real WSDL each message was post fixed with "Request" (for incoming
messages, and "Response" (for outgoing messages).  In the ASMX?WSDL version
all messages were posted fixed with "SoapIn" and "SoapOut" respectively.
- In my original WSDL, the schema was defined in an XSD file.  In the
ASMX?WSDL version the schema was embedded.

Other than that no real difference.  OK, what is the difference in the
outgoing message of the operation that returns an array of the "user"
complexType.

In my XSD here is the relevant parts.

The outgoing message:

<xs:element name="getApplicationUsersResponse">
    <xs:complexType>
        <xs:sequence>
            <xs:element name="users" type="userList" />
        </xs:sequence>
    </xs:complexType>
</xs:element>

complexType userList:

<xs:complexType name="userList">
    <xs:sequence>
        <xs:element name="users" type="user" maxOccurs="unbounded" minOccurs="0" />
    </xs:sequence>
</xs:complexType>

However in the ASMX?WSDL the outgoing message is defined as this:

<s:element name="getApplicationUsersResponse">
    <s:complexType>
        <s:sequence>
            <s:element minOccurs="0" maxOccurs="unbounded" name="users"
type="s0:user" />
        </s:sequence>
    </s:complexType>
</s:element>

I guess this is the difference??  Should my XSD be like the ASMX?WSDL
version of "getApplicationUsersResponse"?  I will do some testing but if
anyone can shed any light that would be appreciated.

Also, is there a guide on controling the output of the ASMX?WSDL somewhere?

Thanks for the help Marvin and Steven.

Thanks,