RE: Andrew Layman and Don Box Analysis of XML Optimization Techniques

Ed Rice wrote:

> I've read the binary xml and we could discuss in detail,

Clarification:   I strongly suspect that Ed means he read and is 
commenting on the work product of the XML Binary Characterization 
Workgroup, not the followup analysis by Andrew Layman and Don Box that's 
referred to in the subject line of this thread.

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------








"Don Box" <dbox@microsoft.com>
Sent by: www-tag-request@w3.org
04/06/2005 06:11 PM

 
        To:     "Rice, Ed \(HP.com\)" <ed.rice@hp.com>, <noah_mendelsohn@us.ibm.com>, 
<www-tag@w3.org>
        cc:     "Andrew Layman" <andrewl@microsoft.com>, "Paul Cotton" 
<pcotton@microsoft.com>
        Subject:        RE: Andrew Layman and Don Box Analysis of XML Optimization Techniques



Ed,

I couldn't agree with your assessment more.  This is why Microsoft is
exploring optimizations that keep us in the text realm, first with
PASwA/MTOM and now with this new proposal that addresses both opaque and
structured data.

DB


-----Original Message-----
From: Rice, Ed (HP.com) [mailto:ed.rice@hp.com] 
Sent: Wednesday, April 06, 2005 2:41 PM
To: noah_mendelsohn@us.ibm.com; www-tag@w3.org
Cc: Andrew Layman; Don Box; Paul Cotton
Subject: RE: Andrew Layman and Don Box Analysis of XML Optimization
Techniques

I've read the binary xml and we could discuss in detail, however in
summary I would say my concerns are as follows;

1) There is nothing which really proves the binary xml format provides
more of a performance impact than doing something like optimization of
the streaming processes.  The Binary is clearly more compact for data
transmissions but opening, validating, compacting/de-compacting could be
quite expensive.  So what's the Net?  Its probably dependant upon the
xml file size and the power of the processor.

2) One of the basic ideas around xml is that it be human readable.
Binary violates this basic premise.

3) Since Binary xml discussed including dictionaries, random access,
random updates to the files, this clearly is more brittle and where an
application receives a partial file it really wouldn't be usable
(instead its just corrupt) as opposed to xml where you have partial
information.

I also have to say, the use cases were very interesting.  However, after
spelling out all the problems with xml to turn and say 'Binary xml would
solve this' without support or explaining seemed a little wrong.

-Ed


-----Original Message-----
From: www-tag-request@w3.org [mailto:www-tag-request@w3.org] On Behalf
Of noah_mendelsohn@us.ibm.com
Sent: Wednesday, April 06, 2005 2:31 PM
To: www-tag@w3.org
Cc: andrewl@microsoft.com; Don Box; Paul Cotton
Subject: Andrew Layman and Don Box Analysis of XML Optimization
Techniques


The TAG has recently given some attention telcons to the question of 
Binary XML and to the conclusions of the XML Binary Characterization 
Working Group [1].  I've recently become aware of an analysis by Andrew 
Layman and Don Box that may be of interest [2].  I believe their work 
supports the following conclusions:

* An alternative to binary XML, Andrew and Don show that a LISP-like
text 
syntax "(...) may optimize better than the existing "<...>" approach.

* As some of us stated repeatedly, detailed experimental measurements
and 
even formal performance models are essential to justifying any
particular 
approach to optimization. 
I suggest we consider these important results in any future discussions.

Noah

[1] http://www.w3.org/2005/04/05-tagmem-minutes.html
[2] http://strongbrains.com/misc/XMLPerf20050401.htm

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Received on Thursday, 7 April 2005 01:48:22 UTC