XML News from Wednesday, July 21, 2004

I've posted the fourth alpha release of XOM 1.0, my free-as-in-speech library for processing XML with Java. While there are still some flaky parts in the internals to be worked out in the next release — there's a nasty bug involving FIXED default attribute values in the internal DTD subset, and the XSLT processing is not completely compatible with the latest versions of Xalan — the API is now considered to be complete and frozen. Code you write to XOM today should not require recompilation against any future 1.x release, and if a really major flaw is discovered in the API design, I'll try to provide a deprecation cycle first before removing the flawed methods. The API has not changed in a backwards incompatible way since 1.0d25. All code that ran with 1.0d25 should not require recompilation to work with 1.0a1. Internal and backwards compatible changes since alpha 1 include:

• Nodes.remove(int) now returns the node removed.
• The IBM virtual machine 1.4.1 is no longer special cased.
• The API documentation has undergone extensive editing.
• The unpublished nu.xom.xerces package has been removed.
• Most methods have been made non-recursive, so they shouldn't cause stack overflows in deep documents.
• The W3C XML Schema Language and WML and HTML DOMs have been removed from the bundled version of Xerces to save space.
• XOM now uses character references only when necessary for all encodings supported by the local virtual machine. However, this may be quite a bit slower than the explicitly supported encodings like UTF-8 and the ISO-8859 character sets. Measurements remain to be performed.
• URI verification and base URI resolution are now performed according to the RFC2396bis algorithm, rather than by using the Xerces and java.net URI classes.
• The Builder class no longer sets any Java system properties for improved compatibility with applets and multiclassloader environments.

I plan one more alpha release to fix the bugs with XSLT processing and FIXED ATTLIST declarations, and then it's on to beta. The beta releases will be pretty much feature complete and bug free, and focus mostly on finishing the documentation.