Interview with Michael Cowlishaw

This article appeared in the August/September 2004 issue of

The Man Behind REXX: z/Journal Visits IBM Fellow Mike Cowlishaw By Gabe Goldberg

Subscribe instantly at www.zjournal.com • Free in the U.S. and Canada • $36 per year outside of the U.S. and Canada The Man Behind REXX z/Journal Visits IBM Fellow Mike Cowlishaw By Gabe Goldberg

n today’s team-oriented world, it’s the Returning to his interest in editors, rare project that runs on one person’s Mike developed Lexx, a specialized (syn- Iinspiration and perspiration. Equally tax-directed color-coding) editor for uncommon are development results that Standard Generalized Markup Language shape technology, affecting how compa- (SGML) text of the Oxford English nies, developers, application users and Dictionary (OED), second edition. Lexx consumers work. and its derivatives are part of IBM’s The REXX programming language to make programming easy.” VisualAge products, and Lexx itself was such a project, conceived and imple- Documented before it was developed recently became available for free down- mented by Mike Cowlishaw, now an IBM and shaped by extensive internal user load. The OED acknowledges Mike Fellow. This title is an executive-level sci- feedback, it quickly grew in popularity twice—for Lexx and editorial contribu- entist/engineer/programmer position, within IBM. Customers learning of it tions; he still consults for the OED. Mike the technical-side pinnacle of IBM’s dual- clamored for its release; it shipped in led IBM’s technical assessment of Java career ladder. The few dozen CEO- 1983’s VM/System Product Release 3, and created NetRexx, a Java platform appointed Fellows pursue self-directed and then was ported to MVS and VSE. REXX version. projects for renewable five-year terms; Now a key resource on all IBM comput- Mike’s books, The REXX Language they serve as corporate consultants and ing platforms, REXX is also available on and The NetRexx Language, and mentors, and report yearly on activities. most other computing platforms. Senior www2.hursley.ibm.com/rexx/ provide Mike joined IBM in 1974, working on consultant/programmer/trainer, Chip extensive REXX information. In this hardware and software products and Davis, remarks, “As soon as I installed interview, Gabe Goldberg, co-editor of interests such as the human-machine REXX on our VM system, all [lan- McGraw-Hill’s The Rexx Handbook, talks interface, programming languages and guages] but Assembler became superflu- with Mike about his eclectic and evolv- editors. Around 1980, he designed and ous. They have been the only two guns ing interests and perspectives on technol- developed REXX, “driven by the desire in my holster for the last 20 years.” ogy and the industry.

26 • z/Journal • August/September 2004 z/Journal: When did you realize the to use. A more modern example is per- huge advance over the scripting and scale on which REXX would succeed? haps Google. macro languages of the day. Nowadays, there are plenty of other languages Mike Cowlishaw: I suppose it was when z/Journal: You’ve articulated the “mini- around that can do the job reasonably the VM development group decided to mal astonishment factor” as a design goal. well, even if they are not as easy to learn. add REXX to CMS before they had even What does this mean and where have shipped EXEC 2. you seen it most violated? z/Journal: Guide us through the tech- nical interests (decimal arithmetic, z/Journal: What would you do differ- MC: I’ve described this in the past as fol- REXX, electronic publishing, ently if you could do REXX and lows: If a feature, accidentally misused, vision/color perception, etc.) listed on NetRexx again? gives apparently unpredictable results, your Web page. Is there a common then it has a high astonishment factor thread among them? One connection is MC: Well, NetRexx really gave me the and is therefore undesirable. This is clear: accelerating REXX decimal arith- chance to correct the things in REXX another way of saying that if the program metic with underlying hardware. that I wanted to correct, such as case- doesn’t match the user’s model of what’s independent comparison and the key- going on, then it has a design error. MC: The common thread is that they are word LOOP for do-loops. all at the “interface” between people and z/Journal: You’ve been quoted as hop- computers, so there is a strong human z/Journal: What’s the most surprising ing that human programmers won’t be element. Decimal arithmetic uses the platform to which REXX has been needed in 20 years. What might lead to base that people use; REXX makes pro- ported? that change? gramming fit the user’s model of the world, not the computer’s; vision and MC: Palm/OS is perhaps the smallest MC: I was perhaps a little optimistic color perception directly affect the way platform that REXX runs on, but really about the 20 years, but I still believe a we present information on displays. none of the ports is surprising. computational approach derived from In all these fields, the biological aspects spiking neuron models will lead to truly mean that the answers are not necessarily z/Journal: REXX has been adopted intelligent systems. Those should be able obvious. For example, from a computing throughout IBM, ported to numerous to handle the tedious side of program- point of view, binary floating-point num- platforms, the subject of dozens of books ming, allowing us to spend more time on bers have the edge over decimal ones—yet and an annual symposium. It’s made sev- the creative aspects. for people, the base just has to be 10. (0.1 is eral programmer generations’ work easi- an infinitely recurring binary fraction, er, more productive and enjoyable. Yet z/Journal: Spiking neuron models? but is trivial in decimal.) it’s not as widely used as some trendy, sometimes ephemeral “languages de MC: Yes, that is a model of a neuron z/Journal: What’s the status of decimal jour.” Why hasn’t the industry adopted where fast learning is seen at the synaps- arithmetic? How do you see it developing? REXX more widely? es; positive if an input spike just pre- cedes an output spike, negative if it just MC: It’s come a long way; the IEEE 754 MC: That’s a difficult question to follows (that is, the connection is standard is being revised to include deci- answer, as any answer must be specula- strengthened if the input is a predictor mal floating-point, and my design is now tion. But I wonder if REXX is perhaps of the output, and vice versa). This fully or partially included in many dif- almost too easy to use; not enough of a behavior has long been suspected, but it ferent languages. And we (IBM) are put- challenge! I’ve described it in the past as has been observed only in real neurons ting it into hardware, too. “BASIC done with hindsight” and in the last 10 years. It’s interesting indeed, some later BASIC variants are because it explains a number of experi- z/Journal: How and when will native- very REXX-like or NetRexx-like. Yet, mental observations, and, I think, will hardware decimal arithmetic be exploit- BASIC still has a bad name in some cir- explain how the brain extracts patterns ed? What effects will it have on cles. I’m confident that many of the best over time (e.g., recognizing words in a programmers and computing? features of REXX (simple syntax, few stream of speech). notations, strong string processing, deci- MC: IBM will be putting it into proces- mal arithmetic, etc.) will continue to z/Journal: How have IBM and the com- sors quite soon. This (and all the related show up in other languages. puting industry changed in your nearly standards) will have a similar effect on 30 years at IBM? decimal arithmetic as the IEEE 754 stan- z/Journal: You’ve indicated that dard had on binary floating-point: it NetRexx is “designed for people, not MC: The biggest change is that most of becomes a no-brainer to use the standard computers.” What other products—com- the easy things—the projects that can be formats and arithmetic in languages, puter-related or not—do you consider to done by just one person—have been rather than inventing new ones. This have been designed that way? done. There are plenty of “big challenges” means we’ll see native decimal formats in left, though (e.g., modeling the brain). most languages, in time, which will make MC: Generally, it’s the products that fol- programming much more pleasant and low an existing, well-known user model. z/Journal: Could you develop REXX quite a bit easier for many applications. One that springs to mind is Lotus and have it blossom in today’s IBM? And they’ll run faster, too. Organizer; this closely follows the “paper-based” organizer model, in both MC: I think so, but probably it wouldn’t z/Journal: What’s planned for future appearance and operation, and so is easy be too successful. At the time, it was a exploration?

28 • z/Journal • August/September 2004 MC: The spiking neural models are a had mentors or role models? Do you and z/Journal: How do you spend your background interest at the moment, as other Fellows mentor IBMers, help orient non-work time? the decimal standards and implementa- them to successful technical careers? tions are keeping me busy. But I’d like to MC: Family holidays are always a good spend more time on that. MC: I don’t think I had a specific men- break, and I also go to Spain for 10 days tor, though I am sure some senior people every summer (without a laptop), and do z/Journal: How do IBM Fellows influ- were helping me along. Nowadays, nothing but hike, explore wild caves, and ence IBM research, technology, and there’s a strong focus on mentoring in generally relax. During the rest of the products? Are Fellows encouraged to IBM; I mentor five people formally and year, it’s mostly work, though I still enjoy focus on probable commercial research about the same number informally. “recreational programming”(such as areas or can you explore where your PMGlobe, a programmable world globe, curiosity leads you? What is work like z/Journal: What changes to IBM and www.cary.demon.co.uk/pmglobe/) and for an IBM Fellow? the industry do you expect in the next electronics (for example, my tiny caving 30 years? lamp). Z MC: There are only 53 IBM Fellows out of the 178,000 technical staff in IBM, so MC: I would expect the variety of tech- we’re spread rather thinly and mostly nical jobs to shrink somewhat; there’s work in very different fields. Those out- no great advantage in having 10 differ- About the Author side research (like me) tend to work clos- ent machine architectures that are only Gabe Goldberg is a er to product development than the a little different, for example. On the technology journalist others, but it varies. other hand, the neural network and enterprise research could well lead to a whole new consultant, focusing on z/Journal: Do you have a staff or do way of doing things. Or there could be Internet and enterprise you work independently? some other breakthrough to keep life computing. A frequent interesting. conference MC: Independently, though last year I speaker/organizer, he had a part-time assistant who helped z/Journal: What would you like to also volunteers with DC-area groups, with verifying the specification for the accomplish technically and professionally? helping professionals and consumers apply IEEE 754 standard. technology to their daily lives. MC: I’d like to be involved in that break- e-Mail: [email protected] z/Journal: During your career, have you through, when it happens.

Subscriptions to z/Journal are FREE in the U.S. and Canada. Subscriptions outside the U.S. and Canada are $36 per year. Click here to subscribe.

z/Journal • August/September 2004 • 29