"BX Proposal"

Here is a proposal for binary XML. Its something I've been thinking about for years and when I saw the call for proposals (the day after the deadline) I thought I'd write it up. Maybe there are some ideas here that will be of use. None of it has been implemented and there are probably many errors.

You will find that it is full of invention. While most of the techniques are well known and many are enshrined in existing standards, the proposal rarely makes use of them. For instance, the "Octet" transform is much like X.690 BER and most of the compression is like bzip2 and gzip. I've done this to present alternative techniques that are tuned to XML specific problems and that take advantage of the specific opportunities that arise, such as (in compression) working on whole symbols for speed, instead of bytes. Otherwise I've tried to avoid invention for its own sake, especially avoiding new data structures that would unnecessarily expose the inner workings of a particular translation mechanism and so fix them into the standard where they might limit future innovation.

I am a retired systems software architect. I worked at Intel for over 20 years on a variety of projects including a couple of standards committees. Now I have my own projects, many of which use XML. Most don't need binary, but 4 or so are much like different use cases documented in the Binary Characteristics report.

My understanding of compression techniques comes mostly from "Managing Gigabytes (2nd Ed.)", Witten, Moffat, Bell, 1999, Morgan Kaufmann.

Attached is the proposal only. I'll follow up with a brief evaluation against the desired properties in a few days.

Paul Pierce
prp@teleport.com

Received on Tuesday, 7 February 2006 13:35:42 UTC