[om] Matlab phrasebook?

Richard Fateman fateman at cs.berkeley.edu
Sun Oct 28 20:02:44 CET 2001



>>
> 
> Here's my take on the subject, my opinion.
> 
> If you're active in this list, it's reasonable to assume that:
> a) you want there to be a standard; and 
> b) you want to make progress in that direction by discussing
> it with people with the same ambition. 
> 
> Richard's original email did not suggest he would subscribe to (a) or (b).


I'm not sure which email this refers to, but assuming that
it has to do with Matlab <--> JavaMath and a CD for Matlab,
I did not think I was attacking the notion of "a standard" nor was I 
trying to hinder the work of people with that ambition.  What I was 
saying was communicating amongst arbitrary programs can be done in a 
variety of ways, and viewing everything as an application of OM is 
probably NOT productive.  It is conceivable, but not in my opinion 
useful, to consider an  OM CD for all sorts of objects and processes. 
Sony Playstations. Microwave ovens. SRS (automobile airbags).  Is this 
comment considered hostile to OM?  Where does one draw the line?  My 
point was:  it is not at all obvious to me that using OM for Matlab 
gains any traction on any problem whatsoever.  Anyone communicating, 
even through OM, to Matlab, would have to be extremely aware of the 
semantics of Matlab.  Any value added by OM  (which, in my opinion has 
yet to be elucidated) would have to be balanced against the downside 
risks, some of which I suggested.
OM (and MathML) are plausible in the CAS <--> browser/display/typesetter
scenario.  Perhaps in the CAS[1]<-->CAS[2] scenario.  I think there is a 
role to play in CAS <-->scientific journals.
So I am not against a standard.

I think that people with the ambition to make their creation into a 
standard have the obligation to listen to, be aware of, and respond to 
criticisms.  In fact a "real" standard requires a public request for 
comments, collecting of comments and responding to them, voting, etc.



> 
> It is perfectly ok to put forward the thesis that the notion of a
> standard is ill-conceived, but I don't think this list is the place to 
> conduct that discussion. 


I'm not attacking the notion of "a standard".  I think though that "the 
OM standard" has problems. Its long gestation period and the 
difficulties apparent to people who have tried to use it support this 
view.  Perhaps I am also concerned by the naivety of the view that OM is 
unique in addressing an important problem and has solved it.
  Techniques for communicating between programs are quite common. Many 
of those techniques are as suitable for communicating mathematics as OM, 
and may be more portable.  Java serialization. Lisp s-expressions. TeX. 
PostScript. even TIFF.  In my opinion, communication between any pair of 
programs can be done directly with less effort than writing 2 
phrasebooks and a CD. You must make your case with n programs, n>>2. 
These programs cannot be too different or else they have nothing in 
common to talk about. For example $p(r+s)$ in TeX is either the product 
of p and the sum r+s, or the function p applied to r+s.  Or maybe the 
probability of at least one of r and s.  It cannot convey which one of 
these is the case... Yet if the programs are too similar (for example, 
two processes running Maple) then OM is just an elaborate distraction.



> By analogy, if you want to question the validity
> of mathematics as a discipline, you wouldn't do that in a mathematics
> journal, but most probably a philosophy journal.



This is not a good analogy.  If I am questioning your design for a 
communication protocol, I am not saying "communication is bad".  I am 
saying your design should be reconsidered.  I say this with a  somewhat 
exasperated tone, but that is because my inherent bluntness is magnified 
by the medium of email to appear rude and offensive. It may also be 
because I had hoped you guys would solve problems of interest to me, 
which seem so central to your goals, but have so far failed.  What problems?
  a. Recall that I have been running a table-lookup integration web site 
for quite a number of years.  Last I looked it was fielding over 120 
queries per day.  Math+database+internet? I've published a few papers on 
this topic.
  b. I have been studying the problem of reading journal articles (OCR) 
containing math formulas, with the intention of studying how to encode 
the results (e.g. TeX, MathML, Lisp, OM?) and how humans can enter math 
into a browser.  I've also published in this area.
c. I actually have programs that read TeX (e.g. the Gradshteyn/Rhyzik 
table of integrals and sums, and produce computer algebra code (Macsyma, 
as it happens) for some 70% of them.  I think this appeared in some 
poster session at ISSAC.
d. I have worked with others who have looked at OM (e.g. for the revised
G&R book published by Academic Press), and found that OM just didn't do 
the job, and did not even seem to be a suitable basis for doing the job.
(Augmented TeX worked.)


A standard has to be judged by whether it is useful. Self-praise is no 
recommendation.

> Most fora
> proceed upon certain basic assumptions which aren't questioned.



I think that if you want to exclude me or anyone else from a discussion, 
you can do so on the basis that I am not sufficiently fundamentalist.
  I can even suggest a name for this discussion:  ClosedMath.


> 
> Richard is welcome to send his thesis directly to me and, I imagine, any 
> of the members of this list, but as individuals. Perhaps he could set 
> up a list in which to debate the virtues of having a standard - I would 
> certainly subscribe. He could even (and perhaps should) publish his thesis
> in a scholarly journal. I just don't think this list is the place to 
> discuss it. 


Actually, I suspect that one or more members of this mailing list have 
suppressed debate, acting as referees, rejecting papers that they viewed 
as not following the party line.
While this is a well-known phenomenon in general, it is one that is 
inconsistent with the advice above.  It would seem to me that advocates 
of OM would welcome hearing possible criticisms here rather than first 
encountering them in public.


> 
> On the other hand, for over a year I've welcomed and given serious
> consideration to Richard's criticisms, read some of his papers and 
> even agreed with some of his points, however, I really think that 
> if this list is to be productive, posts to the list should proceed from 
> the assumption that establishing a standard is a well founded project.



You can make this into a moderated newsgroup, if you feel it would be 
productive to have participants not hear criticisms.  When a 
well-meaning person first set up the moderated Mathematica newsgroup, I 
answered some questions addressed to it  (e.g. Why does Mathematica do 
xyz? ) by responding that Mathematica was designed incorrectly, and that 
one of the consequences was the observed misbehavior, and that other 
systems did the same computation in different ways.  Such comments were 
excluded from the newsgroup on the grounds of irrelevance.  I think the 
newgroup and its moderator have gotten more secure, and now allow 
mention of design issues and other programs.
If OM "fails" whatever that might mean, I doubt that the failure will be 
caused by my distracting comments to this newsgroup.
Regards
Richard Fateman


> 


--
om at openmath.org  -  general discussion on OpenMath
Post public announcements to om-announce at openmath.org
Automatic list maintenance software at majordomo at openmath.org
Mail om-owner at openmath.org for assistance with any problems



More information about the Om mailing list