Yossi Dahan [BizTalk]


Saturday, September 13, 2008

Yet another example for a service's / xml bad design?

One of my clients have been using this 3rd party's software for some time now (I won't name neither organization for obvious reasons).

In their organisation both BizTalk and this software are key components and so they both play part in many scenarios.

Unfortunately, and quite surprisingly, this 3rd party's software doesn't have have any adequate support for integration and the main way to interact with it is pushing some data into 'import tables' in the software's database and then running some EXE to process them. quite horrendous.

To make matters worse this EXE can only allowed to run once at any single time so most implementations require some form of queuing and singleton pattern.

Recently that third party released a version with *some* support for web services, and so, excited by the great news the development team went on to implement a process using the newly introduced web service.

The service that had been exposed relates to the product's reporting features - users can create custom reports using the product's UI and then, using web services, export the result of these queries.

I would have expected, based on past experience, a generic schema in the WSDL that would describe a report's output, irrespective of the fields returned, something like -


<column fieldName=".."></column>

<column fieldName=".."></column>




But the vendor's schema included some generic field from their domain (report name, time of execution, etc.) and then rows and fields based on the specific report generated.

Because the report's structure is not known at design time its fields do not exist in the WSDL, instead a report node exist with an xs:any declaration to include the actual report's contents.

Personally I prefer to avoid xs:any if possible, and I think a schema describing a generic report's result could have been created, but that's not the main problem; the main problem, in my view, is that the fields added underneath the report element (as well as the report element itself), which are generated by the application, all belong to the same namespace.

Because they could not predict the names report designers will use for fields it was more than possible to create an element with duplicate meaning which is a bad idea overall and also causes quite a bit of headache when one needs to create schemas for BizTalk to accurately describe the response of such a service.

One thing I will grant them is that they have though of defining LAX validation for the report, so duplicate elements will not cause validation errors.

Labels: , ,


  • I feel your pain Yossi.

    One of my favourites is the

    <row>value1,value2,value3,value4,valueN</row> implementation. I guess it is the Hybrid model of CSV and XML

    By Anonymous Anonymous, at 20/09/2008, 01:38  

Post a Comment

<< Home