[Coin-standards] moving the offline XML discussion over to the mailing list

[Leonardo B. Lopes]

Great! I think I can start my reply by acknowledging the great offline
work Robin has done in the last month on this project, and also the other
people who have been contributing offline, especially Lloyd Clarke from
ILOG. 

Now moving on to our first technical issue. As I work on refining the
cache in my SP decomposition manager (Some of you saw this project in
Berlin), I realized that in my current XML design, I only considered
performance issues for the way I need to access data. I really wonder what
other people's needs and opinions are on this. First, let me clarify:

By performance issues, I mean the complexity of accessing certain objects 
in certain ways, and even how this complexity should be measured. This
problem is independent of XML, but essential to the standards discussion.
For example, it seems reasonable that obtaining the A matrix for a linear
program should be possible in O(n) where n is the number of nonzeroes. But
maybe the relevant n is the number of columns. The CLP folks might think
the relevant n is the number of rows. 

There are a large number of decisions in terms of data structures which
are dependent on the complexity requirements of commonly performed
operations. Yet, to the best of my knowledge, there is no reference to
what these requirements should be. Maybe it would be interesting for some
of us academics to write a paper spelling out these requirements, by
examining some common algorithms, decomposition techniques,
factorizations, modeling languages, etc... Obviously the context would not
be the complexity of the algorithms themselves, but specifically how
they access outside data. Are there any CLP academics on this list? I
would guess not since most of us are from the Berlin group... Lloyd: Can
you do anything to advertise this list with CLP researchers? Writing this
paper will also help us quantify some of the tradeoffs involved with
using MathML for a subset of the objects. 

So let me start with a concrete example: I think that for optimization
applications, it should be possible to store a matrix as a collection of
blocks with ammendment operations defined on them. Thus, the worst case
performance would be O(n), but the practical performance would be
substantially better. In addition, this approach would provide additional
structure information. Is there something broken with this idea?

Cheers,
Leo.

On Fri, 15 Feb 2002, Robin Lougee-Heimer wrote:

> Leo:
> Since be both agreed that we should move our discussion over to the mailing
> list...here I am.
> 
> Yes, I did talk to Alan King at IBM Research and he expressed interest and
> willingness to participate in the XML standard for exchanging mathematical
> programs (you know,  we need a snappy name).   Alan's currently subscribed
> to the coin-standards mailing list as a digest-member.
> 
> You like the e-optimization idea. A contact person for e-optimization.com
> the  web seminars is Annette Stransky (j-astransky at msn.com). She is a
> marketing consultant for e-optimization.com.
> 
> You asked who to go to for the IOL Practice Portal.  I'm the current
> editor, and would love to promote this effort on that channel (and else
> where on IOL).   IMHO efforts in this direction are of real interest and
> use to the practice community.  For more ideas on IOL channels, ping the
> IOL editor, Matthew Saltzman.  Matt is currently subscribed to the
> coin-standards mailing.  Why not something on the INFORMS E-News
> Announcements?  Also, have you posted a link to your website off Michael
> Trick's OR Webpage (now reincarnated as the INFORMS OR/MS Resource
> Collection  http://www.informs.org/Resources/)
> 
> You sounded positive about my suggestion to write an article for ORMS
> Today.   I'd say even a short news item to let people know about your
> webpage and the mailing list would be good to do.  I had suggested talking
> to Mohan Sodhi, who writes the IOL Cyberspace column.  If you want to write
> an independent article, send a note to Pete Horner (horner at lionhrtpub.com)
> with a short description of your idea.
> 
> I was hoping to announce the phone seminar by Angel Diaz today on  MathML
> and his experience in creating the W3C standard, but it looks like it'll be
> next week.  I think conference calls on subjects related to the standard,
> hosted and run by different companies/people,  would be wonderful way to
> get going.   I hope you're able to do one through Northwestern.  Anyone
> else?
> 
> I've talked to Len Berman, IBM Research, and he's agreed to participate in
> the mailing list.  Len has XML expertise and has work in other standards
> committee.
> 
> Nothing more since Monday on the intellectual property issue  of working
> with standards bodies.
> 
> Brady had kicked off the discussion on some very fundamental questions
> which we need to decide.   For one example of how standard bodies are run,
> check out http://www.oasis-open.org.  Oasis is a not-for-profit consortium
> that designs and develops industry specifications for interoperability
> based on XML.  I've heard that W3C doesn't not do vertical standards. Oasis
> does.  (Note: mathML is under W3C, but this occurred because it was the
> first of its kind).   Participating in Oasis costs money, and provides
> structure.  I'm not suggesting we work under Oasis, I'm just throwing it
> out as one example for those who want background on the way other groups
> have structured standards body.
> 
> more later,
> Robin
> 
> ----------------------------------------------------------------------------------
> 
> Robin Lougee-Heimer
> IBM TJ Watson Research Center
> ph: 914-945-3032   fax: 914-945-3434
> robinlh at us.ibm.com
> http://www.coin-or.org
> 
> 
> _______________________________________________
> Coin-standards mailing list
> Coin-standards at www-124.ibm.com
> http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-standards
> 

========================================================================
Leonardo B. Lopes                                       leo at iems.nwu.edu 
Ph.D. Student                                              (847)491-8470
IEMS - Northwestern University              http://www.iems.nwu.edu/~leo