Some web services serve UTF-8 XML documents with a BOM. Although that practice is disrecommended, it is allowed by the unicode standard. However, java xml parsers have the habit of not reading through the BOM, resulting in an XML wellformedness error.
In kettle the error message (which originates in dom4j) is somewhat cryptic and provides the user no hint that there is an issue with the BOM:
"Content is not allowed in prolog."
The problem can be witnessed with this URL
The attached transformation
PDI-12410.ktr illustrates the error.
Interestingly, the BOM doesn't seem to bother kettle if the source of the XML is a file. Attaching another transformation and XML file + BOM to demonstrate this.