Version: v0.4,  2001-08-12
$Id: README.html,v 1.1 2001/05/01 16:26:40 anderst Exp $

Design

Goals:

Primary

• An binary stream-representation of XML 1.0 information-items,all or a subset.
• As simple stream-representation as possible, KISS.
• Support all XML Schema datatypes
• Use native and IEEE datatypes to facilitate fast stream to/from memory conversions.
• Fast parsing and conversion to in-memory trees, W3C DOM or others.
• Simple and small codebase for encoders /decoders.
• The core BML files should be compatible with Java Micro Edition CLDC 1.0

Secondary
 
• Small binary-stream size compared to XML text-stream-representation, on average anyway. 

However GZIP sizes are a secondary goal compared to performance and ease of use.
• Extensible on in terms of tokens and datatypes.
• Allow encapsulation of OMG IIOP messages in a self describing stream-representation.
• Include small and fast Tree API

Status (v0.40)

• All four representation are now supported.
• An small XML parser has been added.
• Only a few datatypes are currentlu supported but most of these scheduled to be added in the next version.
• The codebase should be really usable when version 0.6 is released .

Changes

TODO

Space considerations

Streamspace saving techniques

• Tag compression
• Native datatypes
• ... there is more in there ...

 

Performance testing

Stream performance enhancements techniques

• Binary tokens
• No need to discover or look for where the next token is.
• No need to handle multiibyte character encodings in tokenization.
• Native datatypes
• No needs to do characters to native datatype conversion, which is a very expensive operation.
• Matches many implementation languages and machines representation of native datatypes

(follows  CDR en/decoding rules without alignment (could be added later)).
• Strings are preceded by a length indicator which improves reading and memory handling.
• Explicit datatyping reduces the amout of syntax and error handling code.
• Relativly small and simple grammer
• Smaller code for parsing and reading.
• Fewer bugs since the codebase is simpler and understandable.
• Produces, on average, smaller streamsize
• Fewer bytes to read.
• Reusable information
• Recurring information may be encoded only once and reused in other parts.
• Namespace handle is simpler since no prefixes are used. Element, attributes and datatypes are all automatically bundled with namespace and localname.
• Some information is packaged in a way that a tree implementation may reuse it without doing extensive / expensive information restructuring and conversion.

Requirements

• JDK 1.3.x    (JDK 1.1.x, JDK 1.2.x should work, not tested though)

• xerces.jar  - for SAX and DOM conversions
• junit.jar    v3.5 - for testing

creates a file named: 'filename.bml'

BML Stream to XML streams:
  java -client org.openebxml.comp.bml.ext.appl.BMLDecompressor -options filename

creates a file named 'filename.xml'

BML Assembler
  java -client org.openebxml.comp.bml.ext.appl.BMLTokens filename

Parses a BML encoded stream and print the tokens to screen.
 
 

Examples

Examples are found in the 'examples' directory
NOTE: Many more examples are scheduled to be created.