Hi Sergey, we analysed thoughrowly the issue and we came to the same conclusion, that the 'action' some how not a content type parameter, but mentioned (escaped) in start-info parameter as part of the type of the main SOAP message part in compound structure.
SOAP with Attachment described in W3C as a feature and the regulatory document shows the principles of how the compound structure have to be built. But there is no proper info how it shall be practically implemented. In CXF implementation in AttachmentSerializer class the implementer references https://tools.ietf.org/html/rfc2387 . But this RFC describes only multipart/related content type. There is no dedicated information about SOAP. We compared behaviour of CXF 3.1.6 with SOAP UI (which we use as a test tool). SOAP UI uses open-ws and the produced SOAP package looks like from older CXF version, means it has 'action' as a property of the main content-type header (proper way). We have seen also the issue in CXF in version 2, which was fixed before 2.7.13, and it seems to be almost the same problem. So I am looking forward to get a response from CXF team. We really need to know is it an issue and will be fixed soon, or it is a change in handling according to some regulations - then we will have to take other steps. Regards, Dmitri >Четверг, 12 января 2017, 19:17 +01:00 от Sergey Beryozkin ><sberyoz...@gmail.com>: > >Hi > >This change introduced an escaping for the 'action': > >https://github.com/apache/cxf/commit/9bbf1acd7291416f9afcd312e7dc302e9031fa25#diff-230a76e41a0370126d1e6e2ea28728eeR170 > >but also in the 2nd Content-Type it is > >type="application/soap+xml; > >Note no closing double quote. > >Effectively, the type is > >type="application/soap+xml; >action=\"urn:ihe:iti:2007:ProvideAndRegisterDocumentSet-b\"" > >where 'action' is a property of the application/soap+xml media type. > >I'm not sure how SOAP w Att compliant it is, just trying to reason why >this change was made... > >Thanks, Sergey > > > >On 12/01/17 15:31, Dmitri Zamyslov wrote: >> Dear Contributors, >> >> during preparation for the migration from Wildfly 8 to 10 we have detected a >> change of CXF behaviour for handling of SOAP with Attachment - >> multipart/related, which caused problems during integration tests. Wildfly 8 >> uses CXF version 2.7.13, Wildly 10 - 3.1.6. The causing difference code is >> located in org.apache.cxf.attachment.AttachmentSerializer and leads to >> different content-type headers: >> >> in Wildfly 8 by sending of SOAP with Attachments: >> >> main content-type header: >> Content-Type: multipart/related; type="application/xop+xml"; >> boundary="uuid:fba1c798-9dc4-4f87-b6a5-dc7534553c80"; start="< >> root.mess...@cxf.apache.org >"; start-info="application/soap+xml"; >> action="urn:ihe:iti:2007:ProvideAndRegisterDocumentSet-b" >> >> content-type header of SOAP part >> Content-Type: application/xop+xml; charset=UTF-8; >> type="application/soap+xml"; >> action="urn:ihe:iti:2007:ProvideAndRegisterDocumentSet-b" >> >> in Wildfly 10: >> >> main content-type header: >> Content-Type: multipart/related; type="application/xop+xml"; >> boundary="uuid:e88e47bb-1abe-4362-b1a4-57a8d9e027dc"; start="< >> root.mess...@cxf.apache.org >"; start-info="application/soap+xml; >> action=\"urn:ihe:iti:2007:ProvideAndRegisterDocumentSet-b\"" >> >> content-type header of SOAP part: >> Content-Type: application/xop+xml; charset=UTF-8; >> type="application/soap+xml; >> action=\"urn:ihe:iti:2007:ProvideAndRegisterDocumentSet-b\"" >> >> There are no other significant differences in HTTP POST messages. >> >> As you can see difference is in the escaped action in content-type header, >> which now is a part of start-info parameter. Because of this difference we >> getting failure from our integration partner: "A header representing a >> Message Addressing Property is not valid and the message can not be >> processed". >> >> Dear contributors, can you please let me know if this change was done by >> failure or it is a right way of handling. If it is a right way please let me >> know which standards describe this behaviour. Thank you very much in advance. >> >> Regards. >> > > >-- >Sergey Beryozkin > >Talend Community Coders >http://coders.talend.com/