My last CDATA fix didn't properly cover the case where CDATA sections occur in sequences. I have created new patches for source and test files to take account of this. I added them to the existing CDATA JIRA ( http://issues.apache.org/jira/browse/TUSCANY-908).
As an aside in making this change I have to do some raw writing to the output XML stream to prevent CDATA markers from being encoded on the way out. I am a little uncomfortable about this as it may lead to XML characters appearing in the wrong place in the output stream. However my change is not the only place that raw writes are used and, from what I have seen, doesn't appear to be consistent. What is the intended mechanism here. Is it SDO's responsibility to ensure that and string type properties are suitably encoded when SDOs are written out to XML or do we intend to rely on uses suitable encoding content before adding to SDO? I would assume the former but it's not clear that this is what the implementation does. It's certainly the case that CDATA changes have an impact here. If we can decide what is intended then we can do a pass of the code making sure it does the correct thing. Simon
