Re: Completion of action item re: XBC

+1.  A nice pair of analyses, and very much to the point.

Quoting from Norm's second note [2]:

> Under the MUST NOT Prevent heading in the table,
> we find that (textual) XML "Prevents" processing
> efficiency, small footprint, space efficiency, and
> forward compatibility. I don't find the properties
> document's discussion of these properties
> convincing on this score.

> From a technical perspective, what concerns me
> most is that the note says "[the] Working Group
> has not done any measurements on submitted
> formats" and yet concludes that Binary XML is
> needed and feasible.

> It may very well be needed and feasible, but I
> don't find the existing documents convincing.

Exactly.  The shame, from my perspective, is that there may indeed be good 
and compelling reasons for a Binary XML, but the work published by XBC 
does not make the case in a convincing and quantitative manner.  I had 
hoped that in the time since the original workshop that more progress 
would be made. 

Perhaps the problem is that, with so many use cases in scope, XBC tries to 
offer something for everyone.  I suspect the case could have been made 
more rigorously if a few key use cases were identified, performance models 
were built (e.g. estimate how many UTF-8 to UTF-16 conversions or string 
comparisons would be required for purposes of interest), and prototype 
implementations carefully tuned and measured quantitatively.  I had hoped 
for statements along the lines of:  "for web services scenarios, binary 
technologies of the following class have been shown to offer 4-7x 
improvement in deserialization of documents of size XXX featuring 
moderately dense markup, resulting in net gains of up to 1.5x for overall 
throughput of representative application servers  and when compared to 
best known text processing techniques".  That would have to be backed with 
all sorts of detail about which documents, which app servers, etc., so 
that one could extrapolate to other similar environments, but if Binary 
XML is worth it then the gains should be easy to spot. 

Like Norm, I'm neither for or against Binary XML.  I do worry that we've 
missed an opportunity to gather the facts that would support a more 
rigorous analysis of the tradeoffs.

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








Norman Walsh <Norman.Walsh@Sun.COM>
Sent by: www-tag-request@w3.org
04/12/05 03:43 PM

 
        To:     www-tag@w3.org
        cc:     (bcc: Noah Mendelsohn/Cambridge/IBM)
        Subject:        Completion of action item re: XBC


I've been carrying an open action item to review the XBC Use Cases and
Characterization documents. I think I've competed them in spirit in
the course of discussions at the telcon, but for the record, I have
completed them in practice now as well:

http://lists.w3.org/Archives/Public/public-xml-binary/2005Apr/0002.html
http://lists.w3.org/Archives/Public/public-xml-binary/2005Apr/0005.html

                                        Be seeing you,
                                          norm

-- 
Norman.Walsh@Sun.COM / XML Standards Architect / Sun Microsystems, Inc.
NOTICE: This email message is for the sole use of the intended
recipient(s) and may contain confidential and privileged information.
Any unauthorized review, use, disclosure or distribution is prohibited.
If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.