close this bookVolume 1: No. 15
View the documentCorrections -- Monash job withdrawn; politics and PYI panels
View the documentNews -- DARPA reorganization
View the documentNews -- Congressional initiatives
View the documentNews -- Alan Salisbury; fuzzy logic at MCC
View the documentNews -- design industry
View the documentDiscussion -- design teams
View the documentDiscussion -- corporate life
View the documentDiscussion -- continuing education
View the documentDiscussion -- promoting yourself
View the documentTools -- newsletters
View the documentTools -- object oriented programming
View the documentDiscussion -- program templates

The Monash position has been withdrawn; no explanation was published.

I said last week that MIT professors would typically be excluded from PYI panels for political reasons. This is more-or- less true for all programs aimed at disadvantaged groups, and for proposals from such individuals, but there is a stronger factor in the case of PYI. NSF is very, very careful about conflict of interest. A professor would not be allowed to comment on any proposal involving his or her school or past advisees -- including students at the school, from the school, or with letters of reference from faculty there. In fact, the professor would have to leave the room whenever such a proposal is discussed. The panel process would be difficult if reviewers from top research schools were invited, so program officers avoid such recruitment unless a specific professor's expertise is absolutely essential. (Even then, the critique might be obtained by mail rather than in panel.) Past PYI winners may be invited to the panel, but they would likely be from the "lesser" schools and under-represented minority groups. If PYI is a beauty contest that favors the top seven schools, it is at least not a positive-feedback loop.

Incidentally, there is very little difference between a bureaucracy and an expert system. If you had to allow for all the factors in complex political situations, your code would be as complex as NSF's rules. NSF's policy manuals aren't terribly hard to understand if you have a week or so to study them. And, interestingly, every rule specifies who has the authority to make an exception.

DARPA has been reorganized, with only the Nuclear Monitoring Research Office unchanged. Although some new clusterings make sense, the overall purpose of the move is unclear. Officially, it is to integrate manufacturing technology throughout the development process in order to reduce costs and speed technology transfer. (The Defense Manufacturing Office has been entirely disbanded or distributed. Could this be continued backlash against Craig Fields' support of HDTV manufacturing research? It would be just like Washington to sow Carthage with salt, or to revise history so that the offending group never existed.) Some observers see a shift from basic research to prototyping, others see it as a shift of applied development funds into basic research areas.

The Information Sciences and Technology Office (ISTO) has spun off its parallel and distributed computing programs to a new Computing Systems Technology Office under Steve Squires, and its machine intelligence and software engineering initiatives to an Intelligent Systems Technology Office [SISTO?] under Barry Boehm.

Other new offices are Advanced Systems (Ron Murphy: aerospace, distributed simulations, mobile target detection, special operations); Defense Science (Lee Buchanan: materials, mathematics, biotechnology); Electronic Systems (Lance Glasser: packaging, displays, etc.); Land Systems (Jim Richardson: armor, anti-armor, vehicle IFF); Microelectronics (Arati Prabhakar: neural networks, optical signal processing, memory technology); and Undersea Warfare (Charlie Stuart: sensors, materials, signal processing, C3I, automation). [Charlotte Adams, Military and Aerospace Electronics, 6/91.]

For more insight into DARPA's AI research thrusts, see the June issue of IEEE Expert. Saul Amarel contributed an overview of the Strategic Computing Initiative (SCI), and Bill Mark and Bob Simpson wrote about the knowledge-based systems program. There are also articles about the Pilot's Associate project and other SCI-funded research. (DARPA has helped bring us a long way from the heyday of COBOL record processing. I just wish I could name a single scientific truth or design principle that we know now and didn't then.)

Senators Hollings and Nunn, along with Jeff Bingaman and other influential committee chairmen, have introduced a package of four bills that meet past White House objections to any government- directed national science policy. Together they call for $1B to be spent by existing agencies, with similar contributions by industry, in support of generic non-defense commercial technologies. Small and medium-sized companies are expected to benefit most. The National Critical Technologies Act of 1991 would coordinate government management of technology and increase monitoring of foreign technology. The Advanced Manufacturing Technology Act of 1991 would direct creation of and funding for a five-year national manufacturing plan. The Manufacturing Strategy Act would increase the Commerce Department's role in developing generic technologies, especially in advanced manufacturing. And the Federal Technology Strategy Act would create a five-year plan for manufacturing, advanced materials, information and communications, and biotechnology. [Brian Robinson, EE Times, 6/24.]

Craig Fields has named Alan Salisbury COO of MCC. Salisbury has been president of Contel Technology Center since 1987, following nearly 30 years with the U.S. Army Signal Corps. He also served on the information and science technology board at DARPA. Salisbury says that MCC doesn't have to "give up long-term results in order to gain shorter-term benefits. The key is to look for intermediate deliverables." [EE Times, 6/17.]

MCC is already exploring AI, neural networks, heterogeneous computing, software technology, and advanced packaging, but they've done almost nothing with fuzzy logic. Now MCC's June 27 "first international industrial Conference on Fuzzy Systems" has brought together leading researchers, including Takeshi Yamakawa, deputy director of Japan's $70M fuzzy logic initiative, and Tomohiro Takage of the Laboratory of International Fuzzy Engineering (LIFE). And Lotfi Zadeh, of course. (Sorry, Lotfi!) Steve O'Hara, the conference chairman, is at (512) 338-3776. [R. Colin Johnson, EE Times, 6/17.]

The Pentagon recently added Flexible Manufacturing to its list of critical technologies, including CAD/CAM/CAE/CAPP, logic synthesis, device simulation, standard product data exchange, database technology and database integration, electronic data interchange, high-speed data transmission, and intelligent interfaces. {Military and Aerospace Electronics, 6/91.]

The National Research Council has published a report on "Mathematical Sciences, Technology, and Economic Competitiveness," $22.50, National Academy Press. (800) 624-6242. [Walter S. Wingo, Design News, 6/17.]

Sandia National Lab continues to have success in parallelizing its 20 most important supercomputer programs. William Camp, manager of the Mathematics and Computational Science Department, says that speedups are typically 10 to 100-fold over vector processing. [Design News, 6/17.]

Brad Rigdon, director of information systems and technology at McDonnell Douglas System Integration Co., heads a government/ industry task force to plan implementation of STEP, an inter- national digital format for product design information. Brad is also president of PDES, Inc., a consortium of 24 engineering firms working on product data exchange in the aerospace, automotive, construction, electronics, and shipbuilding industries. [Walter S. Wingo, Design News, 6/17.] (If you're into concurrent engineering and CIM support, you have to be a team player now. For more on concurrent engineering, see the 7/91 issue of IEEE Spectrum.)

Recent studies say that 20% of all electronic design costs and 37% of product development time are due to design errors. EDA frameworks and design-data management frameworks offer hope of reducing these costs and increasing productivity by 30%. Better-integrated designs may also result. The CAD Framework Initiative (CFI) is currently the best hope for an "open" framework that will permit any CAD tool to work with others in a heterogeneous hardware environment. No vendor currently offers a fully open framework, but DEC's UNIX-based PowerFrame environment is advertised as a good start. [Electronic Design, 6/13.]

Bart Huthwaite (Institute for Competitive Design, Rochester, MI) has some suggestions for developing new products. You may find that they apply to research teams as well, especially if you are proposing to start a research center. 1) Put leadership above technical skill. The leader should be of equal rank with other team members, and must have previous team experience. The group leader must inspire, coach, facilitate, and mediate. 2) Select the team for the task. Don't keep an existing team intact for a new design task. 3) Keep teams small. Three to eight full-time members seem to work well. 4) Develop a mission statement and establish responsibilities before starting the design work. Agree on a set of metrics for success. These metrics should be proposed by the team and approved by management. 5) Give team members the financial support, authority, autonomy, and training to get their job done. 6) Management should ask for periodic reviews, but should avoid direct participation. Individuals (rather than the team as a whole) should be rewarded after the job is done. 7) Appoint successful team members to leadership roles on new teams. [Design News, 6/17.]

Incidentally, the question of team rewards is very controversial. If you reward individuals, there is a tendency for one to sabotage or neglect another in order to look better by comparison. Some experts would recommend rewarding the team as a whole, but not everyone is motivated to achieve an equal slice of a bigger pie. Besides, success is success; how do you assign different rewards to different teams? The best policy appears to be individual rewards based on contribution to team success. The measure should be subjective and the awards somewhat spontaneous -- people will work hard for sincere praise of specific contributions, but become cynical about scheduled awards for meeting management-set performance goals.

The Forestry Service had an interesting "system" that evolved spontaneously. Each employee has the "right" to award a certificate of merit to any other employee -- one certificate per year. Recipients of such awards are naturally quite proud of them. Managers have no part in this, but a wall full of peer praise would certainly have some effect on formal evaluations.

Although more and more households have two working spouses [above 50%, and heading for 75%], the work required of each individual seems to be increasing. Professional work weeks are now often 60, 70, or more hours. Many people (40%, in one Hilton Time Value Survey from Chicago) have less leisure time than a year ago. (Full-time workers average 17.6 leisure hours but want 26.3. Employed women average only 12 hours and want 20.) Almost half of U.S. workers would trade a day's pay for another day off each week. [Carol Kleiman, Chicago Tribune. San Jose Mercury, 6/9.]

Workers who are subject to surreptitious monitoring (e.g., keystroke counting) have increased incidence of colitis, persistent headaches, and nausea, according to a 1985 study of 40,000 women. A 1988 study showed insomnia, chronic fatigue, respiratory illness, and 2.5 times more sick days than before monitoring started. [Leah Krakinowski, SJ Mercury, 6/23.] (These illnesses may afflict you no matter what your source of stress. Bosses and workers tend to have similar levels of stress, but it is the powerless workers who develop the most health problems.)

Accountemps (Menlo Park, CA) surveyed 200 Fortune 1000 VPs and personnel directors, and found that one third would quit if they had start-up capital. [Home-Office Computing, 6/91.] (Perhaps they don't know how little capital it takes. Are they really that impoverished?)

Many stressed-out managers are eager to retire at 55, but perhaps only a third have realistic financial plans. Many plan to expand hobby businesses, and some will lose their homes because of failure to allow for inflation. Cautious retirees will secure their homes and investments before building new careers. [New York Times. SJ Mercury, 6/23.]

Speedup in product design cycles and increasing complexity of products have given engineers more unpaid overtime and more stress. Most are satisfied with their careers -- if they are not in danger of layoff -- although they feel their skills are not used to full advantage. Most don't believe there is a shortage of engineers, except perhaps in specialties like software. (!!!) They would like standardized computer and data interfaces and better CAD/CAE tools to facilitate creative, correct design. Nearly all engineers use PCs for word processing, and administrative tasks often take 10 hours per week. Salaries top out after 15 years or so, and career satisfaction is correlated with attained salary level. For those wanting to get into management, communication skills are essential. [Sherrie Van Tyle, Electronic Design, 6/13.]

Leisure time for Americans has shrunk 37% since 1973, and the average work week has jumped from 41 hours to 47 in that period. Employers can't reduce the hours required, but they should help people work smarter. The U.S. average is 4.5 classroom hours per year; in Japan it is 200 hours. Motorola found a 30-times payback on training; it now invests 2.5% of payroll, and even trains suppliers. IBM workers spend at least 40 hours per year in the classroom, and the success of their Zurich lab helps to draw top scientists. [Alain Beaulieu, Dept. of National Defence, Ottawa. Electronic Design, 6/13.]

Perhaps U.S. engineers get their continuing education from books. Stacey's Bookstore (Palo Alto, CA) reports that the best- selling technical books in EE and CS include C Language Algorithms for Digital Signal Processing; Object-Oriented Design with Applications; C++ Primer; C Programming Language; PostScript Language Reference; and The Power of PenPoint. [Electronic Design, 6/13.]

Then again, why bother with technical training if it's irrelevant to current projects? The knowledge will be out of date in five years anyway. Shouldn't the time be spent on developing managerial or presentation skills?

Not everyone wants to advance, and you may not see self-promotion as necessary. Still, it can help save you from the ax in time of cutbacks, and may save you from unfair action by a new or irrational boss. Marilyn Moats Kennedy, in "Office Politics: Seizing Power, Wielding Clout," suggests that every employee keep a weekly log of accomplishments -- and that every manager encourage the practice. It can be used when writing your self- evaluation prior to formal evaluation by your boss, and also when documenting your worthiness for a raise. Marilyn also suggests the following:

1) After every project or milestone, send a one-page memo to your boss describing the accomplishment. (Save a copy, of course.)

2) Stay on good terms with the secretaries, and make sure they know how productive you are. Secretaries have considerable influence on their bosses. (It is a serious mistake for professional women to cut all ties to lower-ranked women.)

3) Submit press releases to business papers read by your top people.

4) Speak frequently to professional groups. Let your boss know about it, especially if you were asked to speak.

5) Recruit a mentor who will help you become more visible. (It can take six months to two years to set up contacts with a high-level mentor. Research the individual, then join the right associations, work on the right committees, and perhaps play the right sports. Display assertiveness, humor, negotiating ability, and honesty. There is a tendency for men at the top to mentor either women who are exceptionally bright and ambitious or promising young men who are like themselves.)

6) Seek visibility through charity crusades, submissions to the company newsletter, etc. (Such activities also give you an excuse to visit people's offices and chat with them, strengthening your place in the information networks.)

7) Use trade shows to reach competitors. Word may filter back to your own organization about an impressive performance. (Contacts in competing organizations can also be great information sources. You wouldn't discuss trade secrets, of course, but rumors from outside are exceedingly valuable nuggets for your own grapevine.)

The Relayer Group (Scottsdale AZ, (602) 585-3067) sells its AI Trends Newsletter for $295/year, and has been in business since 1984. They also have an AI Directory for $50 and an AI Sourcebook for $500. (All three can be had for $650.)

Larry Press (lpress@isi.edu) mentions some good newsletters at the end of his 2/91 CACM column, and has told me a bit more about them via e-mail. He particularly likes Esther Dyson's Release 1.0, which bridges the gap between the research community and business. Over the last few years she has gotten into expert systems, NLP, groupware, self-organizing systems, e-mail, wide- area information servers, free-text bases, object-oriented programming, grammar checkers, etc., and she often gets to describe interesting products before they come out. Intelligent and well-informed. You can read some of her remarks to the business community in each issue of Forbes. EDventure Holdings, Inc., 375 Park Avenue, New York, NY 10152; (212) 758-3434.

American Programmer is Ed Yourdon's newsletter on software engineering and management (CASE tools, formal methods, software testing, UIMS, etc.). Ed travels extensively and reports cultural notes and analysis with humor. American Programmer, 161 West 86th Street, New York, NY 10024-3411; (212) 769-9460; eyourdon on MCI, or 71250,2322@compuserve.com.

Some others that Larry recommends: Mike Slater's Microprocessor Report, which covers digital hardware, computer architecture, and semiconductor industry analysis. Jeffrey Tarter's Soft newsletter on the business of software -- up on the business trends, and from time to time he "scoops" a new product. Tone Bove and Cheryl Rhoades' newsletter on the multimedia industry. Byte Magazine's weekly newsletter, which leads Infoworld with some interesting analysis from time to time. And for reports on international software development, contact the Computer Software and Services Industry Association (ADAPSO), 1300 North 17th Street, Suite 300, Arlington, VA 22209; (703) 522-5055.

Two others I've seen mentioned are Stewart Alsop's P.C. Letter and David Bowen's Software Success. P.C. Letter is a biweekly that tracks advantages of developing for different computer systems. $345 for members of the Software Entrepreneur's Forum (SEF), $495 otherwise; (415) 592-8880. Software Success is a $197 monthly for developers and software CEOs; (408) 446-2504.

Cobb Group (Louisville, KY) has announced five niche "journals" for software developers. Four are 16-page monthlies (Inside Word, QuickBasic, Inside HyperCard, and DOS Authority) for $59-$99. Turbo C++ is bimonthly for $79. Call (800) 223-8720 for a sample copy. [SJ Mercury, 6/16.]

For a list of object-oriented software development tools -- and addresses of 34 companies producing them -- see Chris Terry's article in the June 6 issue of EDN. There are almost twice as many compilers and tools for C++ as for any other language, and the number (and variety) of class libraries is increasing rapidly. Several major CASE vendors are also incorporating C++ in their development systems.

The growing popularity of object-oriented design is giving the CASE community a chance to rethink their structured design methodologies. Many notations are being championed, and no standards are yet emerging. The 80-member Object Management Group (Framingham, MA) has published an Object Management Architecture Guide, but its contribution is mainly in defining terms and interface requirements.

Objectstore 1.1, an object-oriented database system for most workstations and UNIX environments, will likely be an important C/C++ application programming interface (API). The data manipulation language (DML) supports parametrized types, a mechanism that will be in AT&T's C++ 3.0 and in the ANSI X3J16 standard for C++. $2,000 and up, from Object Design Inc. (Burlington, MA), (617) 270-9797. [EDN, 6/91.]

Asynchronous data communications drivers are available in a $359 C/C++ library from Greenleaf Software (Dallas, TX), (800) 523-9830. Functions include Xmodem, Ymodem, Zmodem, Kermit, and several ASCII protocols, as well as controllers for UARTs, modems, and many communications boards.

We all know the advantages of reusable code libraries, but some design professionals are now arguing that templates are better. A template is a program skeleton, with hooks for calling modules that do domain-specific tasks. Subroutine libraries are easy to maintain and document, but difficult to organize or index and to modify. (The Software Productivity Consortium, Herndon, VA, is working on this problem.) Templates provide a uniform style of user interaction (e.g., Apple's event-driven programming style) and are easy for programmers to use. Only a few general-purpose templates are needed, so the burden of finding and learning an appropriate template is quite small. (In the extreme case, a template becomes an application generator, 4th-generation language, or operating system.)

We need both, actually -- and perhaps expert-system rule bases as well. Templates are useful for controlling complex interactions with many parameters and control paths, such as data entry and validation. (A sufficiently general subroutine would be obese and hard to parametrize. Template customization take the place of knowledge-based compile-time optimization.) As a very simple example, consider how easy it is to modify a template for converting integer vectors to double precision, and how hard it would be to code a generic routine for efficiently converting any runtime-specified vector type to any other. On the other hand, there are times when optimized subroutines will be more efficient than hacked templates. [See John Keller, Military & Aerospace Electronics, 6/91.]