Author: veithen
Date: Sun Oct 13 18:07:46 2013
New Revision: 1531725
URL: http://svn.apache.org/r1531725
Log:
Added an item to the Axiom 1.3 roadmap.
Modified:
webservices/axiom/trunk/src/site/apt/roadmap.apt
Modified: webservices/axiom/trunk/src/site/apt/roadmap.apt
URL:
http://svn.apache.org/viewvc/webservices/axiom/trunk/src/site/apt/roadmap.apt?rev=1531725&r1=1531724&r2=1531725&view=diff
==============================================================================
--- webservices/axiom/trunk/src/site/apt/roadmap.apt (original)
+++ webservices/axiom/trunk/src/site/apt/roadmap.apt Sun Oct 13 18:07:46 2013
@@ -290,4 +290,21 @@ APIs that need to be overhauled
The implementing classes then no longer need to be public (they can be
anonymous or have package access).
In addition, <<<SOAPVersion>>> should be changed from an interface to an
abstract class so that one can
- define static methods to get the SOAP version by envelope namespace or media
type.
\ No newline at end of file
+ define static methods to get the SOAP version by envelope namespace or media
type.
+
+Miscellaneous
+
+* Make non coalescing mode the default
+
+ By default, Axiom configures the underlying parser in coalescing mode. The
reason is purely historical.
+ Axiom originally used Woodstox 3.x and that version implemented one aspect
of the StAX
+ specification incorrectly, namely
{{{http://jira.codehaus.org/browse/WSTX-140}it configured the parser by default
+ in coalescing mode}}, while the specification says otherwise. The problem is
that (poorly
+ written) code that uses Axiom with a parser in coalescing mode doesn't
+ necessarily work with non coalescing mode. Therefore the choice was
+ made to make coalescing mode the default in order to ensure
+ compatibility when using a StAX implementation other than Woodstox 3.x.
+
+ A new major release would be the right moment to change this and make non
coalescing mode the default.
+ This enables a couple of optimizations (e.g. when reading and decoding
base64 from a text node) and
+ ensures that an XML document can be streamed with constant memory, even if
it contains large text nodes.
