XML Performance Improvements through Interdisciplinary Factor Assessment and Application

Don Box's Spoutlet

Syndication

Just over two years ago, I co-authored a position paper on XML and binary encoding techniques that eventually led to our beloved MTOM/XOP combo.
 
Andrew Layman and I have prepared a response to today's XML Binary Characterization that provides a novel approach to the oft-mentioned XML performance problem based on looking beyond the traditional data encoding techniques.
 
You can read our thoughts here.
 
I expect Gudge or I to have an XmlReader/XmlWriter implementation ready in a few weeks. Watch this space.

Posted Apr 02 2005, 03:13 AM by don-box

Comments

Anonymous wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-02-2005 11:21 AM

if you send this to Elliott Rusty Harold, you wouldn't hear from him for another decade.
Patrick Foley wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-03-2005 4:59 AM
Maybe Hunter S. isn't really dead?
MS wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-03-2005 4:03 PM
> Further size reductions are available
> through Adobe ™ PhotoShop ™:

ROTFLA!

This has got to be the funniest piece I have ever read. Good one.
MS wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-03-2005 4:16 PM
> These suggest that the acute presentation of > the left angle bracket in the typical Stag
> and ETag induces turbulent flow at medium
> velocities, and, at sufficiently high
> velocities, shock.

Also, according to Heisenberg's uncertainity priciple, at very high velocities (approaching c), the exact location of tags can't be acertained. No wonder XML guys don't even see the angle brackets anymore -- they just see blonde, brunette, redhead.

http://www.sellsbrothers.com/spout/default.aspx?content=archive.htm#objectvxml
Walter Lounsbery wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-03-2005 4:18 PM
Something bothered me greatly while reading that amazing piece of work on XML improvements. I had encountered this phenomena many, many years ago... in fact, I was strongly reminded of my Master's thesis. You are welcome to add the reference to your material, which can be viewed at:

http://www.lounsbery.com/Lounsbery/Walt/Articles/

Sharp leading angles are indeed bad!

Since the original document was produced on a Xerox Star on 8 inch floppy disk, I've scanned a print into Acrobat format. You will find the related material in Part 4 (page 50 of the original document). I'd give you the title of my thesis here, but my fingers are getting tired... ;-)

Thanks for a great bit of entertainment!
Nicholas Paldino [.NET/C# MVP] wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-04-2005 8:59 AM
The problem with this is that when you send the stream the other way, you actually incur a HUGE performance hit =)
William Stacey wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-05-2005 8:19 AM
However, one can observe that at speeds over Mach 3.5, the beginning angle bracket shows much better dynamics over the open paren for reasons unknown at this time. That said, at those speeds, turbulence at the trailing end becomes the primary factor. Therefore the closing angle bracket with an appended hyphen (e.g.. ">-") shows the least turbulence and greatest speed overall. This also yields an unexpected two factor gain. The hyphen also reduces the sonic boom effect by 70%, and can act as a little rudder in the back. We are still studying the effect of “-<” at the head and adding Canards on both sides. We feel this will give it a coolness factor (like a rocket or an F16.) Will post results shortly.
Steven DeRose wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-06-2005 6:04 PM
Brilliant research breakthrough! I did, however, notice two areas in which you might seek further improvements:

First, drag can be radically reduced by avoiding discontinuities in the cross-sections of successive slices. Thus, instead of

<p> The World Wide Web Consortium</p>

you could achieve laminar flow by transmitting:

<p style="text-decoration: underline overline">The World Wide Web Consortium</p>

Second, at sufficient speeds, a properly-shaped projectile can create a "supercavitation" bubble entirely surrounding itself, thus travelling in a gaseous space even while submerged in a liquid. This reduces drag by about 1000x, as in the Russian VA-111 Shkval torpedo, which operates at more than 230mph.

The optimal shape for supercavitation is a flat, sharp-edged nose. The XML "<" is thus ideal. But supercavitation is not achieved for XML transmissions in practice. It is likely the reason for this is that the present Internet is incapable of accelerating XML data to the minimum velocity required, about 100mph.

Should this general Internet limitation be overcome, XML documents will naturally begin to supercavitate, making them nearly 1,000 times more efficient than other documents.
Dave Bacher wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-07-2005 11:14 AM
How does it stand up to a crash test?

If your application hits a UAE at 5, 10, 25mph, how does it survive? Are the passengers safe? Does the crumple zone protect enough data for someone to recover it?

What about side impact, the sides of that fragment look awful weak to me, maybe you better encapsulate every character and give it a namespace to pad it in case of a problem:
<p style="text-decoration: underline overline"><char encoding="us-ascii">T</char><char encoding="us-ascii">h</char><char encoding="us-ascii">e</char><char encoding="us-ascii"> </char><char encoding="us-ascii">W</char><char encoding="us-ascii">o</char><char encoding="us-ascii">r</char><char encoding="us-ascii">l</char><char encoding="us-ascii">d</char><char encoding="us-ascii"> </char><char encoding="us-ascii">W</char><char encoding="us-ascii">i</char><char encoding="us-ascii">d</char><char encoding="us-ascii">e</char><char encoding="us-ascii"> </char><char encoding="us-ascii">W</char><char encoding="us-ascii">e</char><char encoding="us-ascii">b</char><char encoding="us-ascii"> </char><char encoding="us-ascii">C</char><char encoding="us-ascii">o</char><char encoding="us-ascii">n</char><char encoding="us-ascii">s</char><char encoding="us-ascii">o</char><char encoding="us-ascii">r</char><char encoding="us-ascii">t</char><char encoding="us-ascii">i</char><char encoding="us-ascii">u</char><char encoding="us-ascii">m</char></p>

Of course, you'd have to include (at a minimum) UTF-8, UTF-7, Big5, Shift-JIS, UTF-16, Klingon and Smurf for each character (in a production environment).

Of course, SQL server will require additional attributes on the char object, which will be in a really long namespace, which will be aliased to the entire name of the namespace. It is recommended that the namespace be at least 1000 random Unicode characters, drawn from the full system, to reduce the chance of a collision. Additionally, each character will require two GUID numbers to additionally identify it.

mario wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 04-12-2005 7:08 PM
very interesting thoughts.

But in my mind, it's better to use heating theory for system development description in natur. And, if I'm right,the described equation of reynolds/stockes should only be use in linear environment. If you have turbulences the used eq. switches into non linear model and we need the some bigger equations, format A3 landscape. All the automotive wind channel work is based on non linear models and they need calc. power without any end.

That means, for internet and other process driven env. seems better to use thermodynamics with time depenency. Entropie is than a genious number to describe system stability. I did not write such a nice article, but you can refer to ISBN: 3-458-33885-3 Ilya Prigogine: the laws of chaos, but this book is in german.

So let us modify our article the results are interesting enough. Think on a SOA with processes/ services, which are stable?? How stable is the first BP design?
regards
mario
mail:
mb@consult-biedermann.com
ps hope not to loose my physics degree??
Andrew Layman wrote re: XML Performance Improvements through Interdisciplinary Factor Assessment and Application
on 05-07-2005 7:19 PM
These suggestions, so modestly posted, indicate a depth and breadth of knowledge, offering tantalizing possibilites for greatly improving my crude work. I thank you all.

Add a Comment

(required)  
(optional)
(required)  
Remember Me?