Re: "Binary XML" proposals

[Al Snell <alaric@alaric-snell.com>]

On Tue, 10 Apr 2001, Sean McGrath wrote:

> You gotta run your figures on this sort of thing. If I had a days
> holidays for every week I have wasted on an "obvious" performance
> improvement over the last 18 years, I would not be seen on
> XML-DEV till next April fools day.

But before we can benchmark my idea for a binary encoding of XML vs. XML
in real world applications, we need it implemented. And to decide whether
it's worth implementing, we need to make some theoretical estimates first!

Indeed, we can't assume that it will help the average application, but it
may help in the use of XML for things like image storage,
communications-latency-criticial systems, systems with limited parsing
speed, etc - a not inconsiderable section of the market for data storage.

The tree data model of XML is well suited to many tasks that the textual
encoding is not. Heavily numeric data such as images is a big class of
applications that are missed out. A scientific data set of a billion
floating point numbers can fit into 4 bytes per element in a binary
array (depending on the accuracy of float you use), or:

<s>0.39572756492383E-12</s>

...in textual XML (I think that's about the number of sig figs you get in
a 32 bit IEEE float... very roughly). Even in my self-describing scheme
that would be 5 bytes per element (one tag byte saying "HERE BE AN
IEEE FLOAT", then 4 bytes of float[0]).

Storing that as textual XML would be horrendous... and the
notepad-readability of XML wouldn't be incredibly useful in such a sea of
data anyway :-)

> 
> regards,
> Sean
> 

ABS

[0]My book on IEEE floats is at home, so don't quote me on that.

-- 
                               Alaric B. Snell
 http://www.alaric-snell.com/  http://RFC.net/  http://www.warhead.org.uk/
   Any sufficiently advanced technology can be emulated in software