Path: utzoo!utgpu!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!pasteur!ames!mailrus!iuvax!ndcheg!jeff From: jeff@ndcheg.cheg.nd.edu (Jeff Kantor) Newsgroups: comp.sys.mac Subject: Mathematica - I like it ( a lot), but beware of the hype Keywords: Mathematica, symbolic mathematics Message-ID: <601@ndcheg.cheg.nd.edu> Date: 28 Jul 88 21:30:12 GMT Organization: Dep't of Chemical Eng., Univ. of Notre Dame Lines: 105 Yesterday I received my copy of Mathematica, and thought I would post a few comments. As a sometimes user of SMP (a Symbolic Manipulation Program) developed about 5-7 years ago by Steven Wolfram, I was attracted to Mathematica as a possibly friendlier and updated variant that would run on a less expensive Macintosh platform. I need such a tool to carry out routine, but rather lengthy and error prone symbolic calculations in the area of nonlinear control theory. I ordered the Enhanced version for the Mac II. (list price: $795). I have successfully loaded and run this on a Mac SE with 4MB of memory and the Radius Accelerator w/68881. Now that I have had a chance to browse the manual, apparently the only difference between the Standard version (list price $495) and the expensive version is support of the 68881. Frankly, $300 seems like a very steep price to pay for 68881 support. Incidently, there is no upgrade policy. If you have the standard version, and later decide for the enhanced version, be ready to fork over the full price for the new version. This bit of extortion is really uncalled for, in my opinion. I wonder what the upgrade policy will be when version 2.0 comes out? As others have already noted on the net, the Mathematica book from Addison-Wesley contains the how-to-do-it bit. You must purchase that seperately in order to find out how to 'do' Mathematica. The manuals that come with the software only describe the Macintosh front-end. It's sort of like the case for TeX, where the TeXbook is a vital part of the documentation. I think it's a bit cheesy and presumptuous not to include the $30 paperback with the software, but there you are. The software is readily installed using the enclosed instructions. It takes about 3MB of disk. It takes a minimum of 2MB ram to load the front-end and kernal. It's better to have 3MB for problem solving room. 4MB is probably more realistic for any kind of serious use. Perhaps more. Before blasting Wolfram Research, however, consider that SMP regularly uses 10-20MB on a Sun Workstation (virtual memory saves the day.) Symbolic manipulators require elbow room (Maple notwithstanding). So that is simply what it takes to put big software in small workstations. Moreover, if the kernal is running remotely, the user can get by with 1MB running only the front-end. The catch is that not very many of us yet have access to a remote kernal. I can only imagine what will happen to campus computer centers when Calc 1 students all start up remote Mathematica kernals the night before problem sets are due! The front-end is very appealling. The notebook metaphor is very nice, I'm tempted to write a sort of 'living paper' regarding nonlinear observer design for chemical reactors. Illustrations, lot's of graphics, see what happens when certain models are introduced for the reaction rate expressions, do the closed loop simulations (numerically), that type of thing. It's worth the $495, and may well be a revolutionary product in its own right. I don't say that lightly. The front end has some sort of postscript interpreter builtin. You can type PostScript commands, tag them with a PostScript attribute, and presto, format the graphic. And it goes two ways, unformat a graphic and you see the PostScript description. The manual says that its not a complete PostScript Implementation, but doesn't say what's not there. The graphics are pretty outrageous. Shading, lighting, parametric plotting, etc. Also animation, provided the individual frames are computed ahead of time. The grey levels look OK on the Mac screen, and actively good on the Laserwriter output. Can't wait to see how things look in 256 colors that are blended together by some sort of dithering algorithm. The Mathematica kernal will seem very familiar to SMP users. The syntax is cryptic. It will probably be flamed by a lot of neophytes, but praised by SMP users as a significant improvement. In fact, I like Mathematica. It's brief and to the point, the control structures are not too hard to figure out, and its suited to doing Math computations. But it is complex, just take a look at Appendix B of the Mathematica book to get an idea of all the builtin functions. Learning to program in Mathematica would be like learning to program in any rich high level language. One should expect a learning curve, maybe a few days to learn basic common commands, a few weeks for writing functions, maybe several months to become really proficient. The online help is very well done. I'm sort of old-fashioned, I don't esp. care to read and contemplate complex matters written on a computer screen. But others may find help to be an effective substitute for the book. The software software has a nice feel to it, seems speedy enough and reasonably robust. Spelling errors in the dialog screens make me a bit nervous about where other bugs might be. So is this a revolutionary product like the newspapers say? For scientific and engineering work, the Mathematica kernal is a practical alternative to mainframe symbolic algebra packages. Mathematica seems to be good implementation for a low end workstation environment. So the kernal should be regarded as an evolutionary product. By itself, something worthwhile. The Front-end provides an exceptionally fine way of using the Mathematica kernal. The ability to write 'living papers' has much potential for both education and mathematically oriented engineering research. Much more so, in my opinion, than so-called hypertext. In summary, the overall Mathematica environment (kernal and front-end) might well change how I do my analytical work. I'll spend the time it takes to learn to use it well. It will nicely complement my other main research tool, Matlab, a programmable numerical linear algebra package. But I wonder if people other than technical professionals will be willing to pay the stiff price, or tolerate the arrogance of Wolfram Research sales policies. Jeffrey Kantor Univ. of Notre Dame jeff@ndcheg.cheg.nd.edu