• Secure C ... Not a Moment Too Soon!
• Upgrading DACS' Microsoft Visual Studio Support
• DDC-I In Action - Sikorsky's S/H-92: Leading the Charge in Safety & Standardization
• Thoughts from Thorkil - Floating Point Concepts and the 80x86 Implementation (2)
• Unusual Solutions - Part 2

By David Mosley

DDC-I has a long history supporting safety-critical software and has been a strong proponent of Ada for that very reason. The Ada vendors have, unfortunately an aspect of Don Quixote tilting at the windmills of C and C++. Perhaps the world of C is finally ready to listen.

Thomas Plum (of Plum Hall fame) reports in the latest issue of SD Times (February 15, 2005 issue - Letters to the Editor ...An Orthogonal View of Software Monocultures - http://68.236.189.240/letters/index.html) that the committees that work on the C international standard are developing a Secure C Library. 

Quoting from the technical report, "This technical report provides alternative functions for the C library that promote safer, more secure programming. The functions verify that output buffers are large enough for the intended result and return a failure indicator if they are not. Data is never written past the end of an array."

Considering that the basis for many computer viruses which allows them to do their damage and propagate themselves, is an unanticipated buffer overflow, then adoption and use of Secure C cannot come too soon. Plum also notes that the compiler will also be changed so that every fetch-or-store operation upon an array will require adequate bounds and type information to ensure that undefined behaviors do not take place. 

Of course, bounds checking has been a part of Ada since the earliest drafts. Imitation is still the sincerest form of flattery. For more information about the Secure C Library see the latest draft at www.open-std.org/JTC1/SC22/WG14/www/docs/n1088.pdf.

[ Back to Top ]

Upgrading DACS' Microsoft Visual Studio Support

According to DDC-I Senior Software Engineer Richard Frost, the company's proven DACS product has always offered support for objects created by Microsoft Visual Studio. A conversion utility was used to "fix up" an object file for compatibility with the DACS linker.

Unfortunately, later releases of Visual Studio - including .Net - have added incompatabilities in the LIBC library, as well as to the initialization code for the Run Time System (RTS).

"Addressing the increasing issues experienced by a growing customer base using a combination of development tools hosted on Windows became a priority," explains Frost, who's Windows experience includes leading the rehosting program for DDC-I's TADS product line.

With the DACS 4.7.16 release, DDC-I now provides two versions of the DACS RTS. The default product is built with the standard DACS assembler and fully compatible with the standard target linker provided with DACS. An optional RTS, built with Microsoft Visual Studio, is available to every DACS customer.

Due to legal limitations, DDC-I is not able to ship the "link.exe" component of Visual Studio. Instead, a template file is provided which calls the linker and uses the Microsoft compiled version of the DACS RTS, eliminating incompatabilities and providing the customer an easy integration path for Microsoft Visual Studio and .Net.

[ Back to Top ]

DDC-I In Action

Sikorsky's S/H-92: Leading the Charge in Safety & Standardization

Process and product standardization takes flight during S/H-92 development using DDC-I tools

The S/H-92 medium lift twin-engine helicopter, evolved from Sikorsky's legendary S-70 Black Hawk and Seahawk. When it officially took to the air, FAA New England regional administrator Amy Corbett said, "On December 17 [2002], precisely 99 years after the Wright brothers' flight at Kitty Hawk, the S-92's certification was signed. What a wonder it was to see the birth of the world's safest helicopter, on the same day as the first flight."

Where the Wright's first shaky experiments seem risky by today's standards, the "new generation" S/H-92 is designed around the concept of safety. Based on technology perfected during over five million flight hours of global mission experience logged by Black Hawk and Seahawk, it is the first helicopter in the world certified by the European Aviation Safety Agency/Joint Aviation Authorities (EASA/JAA) to the latest, most rigorous safety standards. The S/H-92 is also the first certified by the FAA to FAR Part 29 Transport Rotorcraft, Amendment 47, the latest U.S. safety regulations. 

Beating deep within the digital heart of the next-generation flight-control and fuel- control avionics is safety-critical code generated with proven DDC-I software development tools. Primary contractor Sikorsky used DDC-I tools to develop redundant flight control systems code as S-92 software development came online in the mid-nineties. Hamilton-Sundstrand, working for engine subcontractor General Electric, coded the FADEC (Full Authority Digital Electronic Control) engine and fuel control systems. Code originally developed for the European NH-90 "Eurocopter" project was wisely (intentionally) re-used and updated during development.

A December 1998 maiden flight at the Sikorsky Development Flight Center in West Palm Beach, Florida, preceded a production launch in 1999. First orders landed in 2000. Working with Sikorsky in project development was an international consortium: Embraer of Brazil, Gamesa of Spain, Mitsubishi Heavy Industries of Japan, Jingdezhen Helicopter Group/CATIC of the People's Republic of China, Aerospace Industrial Development Corporation (AIDC) of Taiwan and major subcontractors Hamilton-Sundstrand, Rockwell Collins and General Electric.

Also noteworthy in the S/H-92 is Sikorsky's consideration of intelligent standardization of technology between civil and military development. The desire to reduce the cost of avionics and produce a more versatile, universal S-92 "has caused... a market conversion," according to Sikorsky manager Mike Myers, reflecting on the early development strategy of Sikorsky and Rockwell Collins to build systems certifiable in both commercial and military markets, due to demand from military customers for H-92 avionics also certified for civil operations.

The commercial S-92 is designed for roles such as passenger transport for the offshore oil and airline industries, executive transport and civilian Search & Rescue. The military H-92 Superhawk prefers jobs like utility and troop transport, anti-submarine warfare, combat Search & Rescue and Head-of-State transport. Advanced rotor systems and 5000+ horsepower from dual General Electric CT7-8A turboshaft engines carry 19 passengers in airliner comfort 490 nm (907 km), search for a full hour 275 nm (510 km) from base to rescue six survivors and return to base -- with 30 minutes of reserve fuel -- and perform at weights to 31,000 lb. (14,065 kg).

Intelligence extends throughout the helicopter, as the S/H-92 monitors and captures all operational airframe and flight data and identifies system health issues in real time, providing an onboard display to enhance flight safety and an integrated air-to-ground Configuration and Maintenance Management System (CMMS) for smart maintenance and -- that's right -- increased safety. Mission flexibility and low operating costs were primary design points at program inception, aimed at providing a multi-mission helicopter with unprecedented levels of reliability, maintainability and operational economics.

At 7:30 a.m. on February 21, 2005, a Norwegian sunrise met the first Sikorsky's S-92TM entering revenue service over the North Sea, nineteen Statoil employees in full survivor gear taking part in the historic first flight. The S-92 flew the first of four flawless round trips, transporting a total of 152 different passengers to and from various Statoil platforms in the North Sea. Operated by Norsk Helikopter on contract to Statoil, Norsk is the first European off-shore oil S-92 customer. A second S-92 is scheduled to enter service in spring 2005, and Sikorsky anticipates several S-92's will begin offshore operations in the North Sea within the year.

[ Back to Top ]

Floating Point Concepts and the 80x86 Implementation (2)

By Thorkil B. Rassmussen 

Here we shall take a closer look at the Ada model numbers, the problems with sharing the NPX between Ada tasks, use of NPX inside interrupt handlers, and how to employ the NPX for maximum performance.

The model numbers are introduced in the Ada83 Reference Manual (ARM) in sections 3.5.6 and 3.5.7, and put some constraints on the number of significant bits (B) with respect to the maximum exponent, expressed as a power of 2. The binary canonical form is

sign * mantissa * (2 ** exponent)

where sign is +1 or -1, and mantissa is a value of the form 0.1xxxx... (except when the value is zero).

The model numbers comprise all numbers where the number of significant binary digits B obeys the constraint that the exponent is in the range -4 * B .. 4 * B.

For 32 bit floats there are maximum 24 mantissa bits, requiring a binary exponent range from -96 .. +96, and for 64 bit floats there are maximum 53 mantissa bits, requiring a binary exponent range from -212..+212, both well within the maximum exponent range for the float type.

Numbers are not model numbers when the exponent exceeds the 4 * B range, but can still be within the safe numbers with a maximum exponent E limited by an implementation defined value larger than 4 * B; for 32 bit floats this value E is 125 and for 64 bit floats 1021.

The model numbers intend to provide a predictable accuracy when doing floating operations, where the operands are (each) within a model number interval, and the result within the interval defined by the operand intervals applied to the operations.

The benefit of the model numbers is more questionable, as floating point processors live up to the requirements for model numbers (ARM section 4.5.7), so an application need rarely be designed in terms of model numbers. Floating point constants are always expressed as safe numbers and rounded to the value of B in question and allowing a larger exponent than 4 * B. A comparison of a float calculation to a float constant should therefore consider that the result is not necessarily exact, but may be in the model number interval around the float constant.

Using the 80x86 NPX processor for floating point calculations

The 80x86 Ada compiler will generate floating point instructions according to the float expressions required, but there are some significant issues that must be considered. In a previous article it was explained how the WAIT instructions must be used to assure synchronization between the floating point coprocessor and the > main integer= processor, so that a result floating value is in fact in memory, when the main processor needs it.

A further complication is tasking. When a task relinquishes control asynchronously, e.g. as a result of preemption, it may or may not have floating point processing undergoing. The floating point state must therefore be preserved as well as restored when the task is resumed. This may be a time consuming operation as it involves a lot of floating point information that needs to be saved and restored. For a 32 bit processor there are 108 bytes to handle. Since presumably not many tasks actually do floating point operations this is just slowing down the task switch operation.

Therefore the Ada RunTime System (RTS) employs a lazy float handling strategy. The first time a task executes a float operation it will trap through interrupt 7. This trap occurs because each task switch operation enables the trapping situation by setting the TS bit in the processor=s CR0 register. The only tasks that can be hit by interrupt 7 are then those tasks that execute floating point operations after they have gained control from another task.

The RTS therefore keeps track of which task is the current owner of the floating point processor (NPX). There may be no owner. If a task traps via interrupt 7 it has executed a float operation, and there are several possibilities:

1. The task may be the current owner of the NPX, and can then just remove the TS bit in CR0 and redo the float operation, effectively resuming from last time the task was active
   

2. There is no owner of the NPX - here there are two possibilities:
   - This is this task= s first NPX operation, so a fresh NPX state is loaded
   - Restore a previously saved NPX state
   In either case remove the TS bit in CR0 and mark this task as the current owner and resume the operation again.
   

3. Another task is owner - the NPX state must be saved for this other task first, and then the handling is as if there were no owner.

With this scenario only tasks that do floating point operations pay for saving and restoring the NPX, and other tasks can be switched to without any extra overhead.

A special issue is executing floating point operations within an interrupt. The interrupt routine is running > on the back= of the otherwise executing task which could be the NULL task. This task could already be doing floating point operations at the point of interrupt. Special RTS calls are required by the interrupt routine to ensure that the current owner of the NPX is saved and the owner switched to the interrupt code. Also, leaving the interrupt routine a similar RTS call is required to assure that there is no owner of the NPX anymore. We cannot use the interrupted task directly, unless its possible use of the NPX has been saved first. More importantly, when this task eventually resumes it should continue with its proper NPX. Therefore it is vital that interrupt code is not again interrupted, while the NPX owner is > fake= .

In some contexts interrupts are unconditionally enabled (e.g. when entering exception handling), so that poses an extra problem that must be handled in the RTS for exception management. But in general interrupt routines should be disabled for as little as possible and do their potential floating point operations within an NPX-safe region. Allowing exceptions during the floating point operations must take the necessary precautions.

Optimal use of the 80x86 NPX floating point processor

The 80x86 NPX works in stages, and parallelism is achieved by assuring that each stage has something to do. A stage can be to load a float value into the processor, to perform an operation on already loaded value, or to store a result to memory.

To make a simple add of two floats: A := B + C; will have to await the loading of B and C, the plus operation and the store to A, in all 4 stages are required. There is not much to do about that, except when such occurs in loops, where each loop iteration is independent of the other:

-- a, b and c are float arrays:
for i in 1..12 loop
  a(i) := a(i) * b(i) + c(i);
end loop;

i:= 1;
loop
   a(i) := a(i) * b(i) + c(i);
   a(i+1) := a(i+1) * b(i+1) + c(i+1);
   a(i+2) := a(i+2) * b(i+2) + c(i+2);
   i := i + 3;
   exit when i > 12;
end loop;

--------------------------------------------------------------------------
7: loop
8: a(i) := a(i) * b(i) + c(i);
--------------------------------------------------------------------------
0000002D   8BB560FFFFFF    MOV   ESI,[EBP-160]
00000033   D944B5C0        FLD   dword ptr [EBP+ESI*4-64]
00000037   D84CB590        FMUL  dword ptr [EBP+ESI*4-112]
0000003B   D884B560FFFFFF  FADD  dword ptr [EBP+ESI*4-160]
00000042   D95CB5C0        FSTP  dword ptr [EBP+ESI*4-64]
--------------------------------------------------------------------------
9: a(i+1) := a(i+1) * b(i+1) + c(i+1);
--------------------------------------------------------------------------
00000046   D944B5C4        FLD   dword ptr [EBP+ESI*4-60]
0000004A   D84CB594        FMUL  dword ptr [EBP+ESI*4-108]
0000004E   D884B564FFFFFF  FADD  dword ptr [EBP+ESI*4-156]
00000055   D95CB5C4        FSTP  dword ptr [EBP+ESI*4-60]
--------------------------------------------------------------------------
10: a(i+2) := a(i+2) * b(i+2) + c(i+2);
--------------------------------------------------------------------------
00000059   D944B5C8        FLD  dword ptr [EBP+ESI*4-56]
0000005D   D84CB598        FMUL dword ptr [EBP+ESI*4-104]
00000061   D884B568FFFFFF  FADD dword ptr [EBP+ESI*4-152]
00000068   D95CB5C8        FSTP dword ptr [EBP+ESI*4-56]
--------------------------------------------------------------------------
11: i := i + 3;
--------------------------------------------------------------------------
0000006C   838560FFFFFF03  ADD  dword ptr [EBP-160],3
--------------------------------------------------------------------------
12: exit when i > 12;
After optimization for speed is applied:
--------------------------------------------------------------------------
7: loop
8: a(i) := a(i) * b(i) + c(i);
--------------------------------------------------------------------------
0000002D   8BB560FFFFFF    MOV  ESI,[EBP-160]
00000033   D944B5C0        FLD  dword ptr [EBP+ESI*4-64]
00000037   D84CB590        FMUL dword ptr [EBP+ESI*4-112]
0000003B   D944B5C4        FLD  dword ptr [EBP+ESI*4-60]
0000003F   D84CB594        FMUL dword ptr [EBP+ESI*4-108]
00000043   D9C9            FXCH ST(1)
00000045   D884B560FFFFFF  FADD dword ptr [EBP+ESI*4-160]
0000004C   D944B5C8        FLD  dword ptr [EBP+ESI*4-56]
00000050   D9CA            FXCH ST(2)
00000052   D884B564FFFFFF  FADD dword ptr [EBP+ESI*4-156]
00000059   D9C9            FXCH ST(1)
0000005B   D95CB5C0        FSTP dword ptr [EBP+ESI*4-64]
--------------------------------------------------------------------------
9: a(i+1) := a(i+1) * b(i+1) + c(i+1);
--------------------------------------------------------------------------
0000005F   D9C9            FXCH ST(1)
00000061   D84CB598        FMUL dword ptr [EBP+ESI*4-104]
00000065   D9C9            FXCH ST(1)
00000067   D95CB5C4        FSTP dword ptr [EBP+ESI*4-60]
--------------------------------------------------------------------------
10: a(i+2) := a(i+2) * b(i+2) + c(i+2);
--------------------------------------------------------------------------
0000006B   D884B568FFFFFF  FADD dword ptr [EBP+ESI*4-152]
00000072   D95CB5C8        FSTP dword ptr [EBP+ESI*4-56]
--------------------------------------------------------------------------
11: i := i + 3;
--------------------------------------------------------------------------
00000076   83C603 ADD ESI,3
--------------------------------------------------------------------------
12: exit when i > 12;
--------------------------------------------------------------------------
00000079   89B560FFFFFF    MOV [EBP-160],ESI
0000007F   83FE0C          CMP ESI,12
00000082   9B              WAIT
00000083   7F02            JG 00000087H

It is important that only /optimize=speed is used and not the general /optimize, unless the common subexpression (cse) optimization is excluded explicitly (/optimize=all=on,cse=off,speed=on), since cse-optimization may counteract the optimization for speed.

It is however not always easy to rewrite float expressions, but often worthwhile as the example shows.

About the Author Thorkil Bjørn Rassmussen has worked with DDC-I for over 20 years. He has a Master of Science, Computer Science, from University of Copenhagen. Thorkil has substantial experience with all of the DACS tools and is the key individual involved in all FAA certifications for the DACS product line. Thorkil lives with his wife Jane and two children Jonas and Tine, just outside of Copenhagen, Denmark.

[ Back to Top ]

Unusual Solutions - Part 2

By Linda Rising
risingl@acm.org
www.lindarising.org

Last month (http://www.ddci.com/news_vol6num2.shtml), I described the unusual solutions of survivors in Sri Lanka as they faced the terrible tsunami in Southeast Asia. I think in any difficult situation, there are those who conquer the complexity with intriguing approaches. We can all learn from these exceptional individuals (or animals).

In this article, I’m going to tell you about a trip I took to New Mexico a few weeks ago. No—no tsunami hit that state, but New Mexico is like many places that face a list of serious problems. Here are some:

(1) They have an increasing number of disadvantaged students who are smart and might be willing to pursue college programs, but they face almost insurmountable obstacles such as: lack of family support, single parent status, no viable source of income, no encouragement from peers, inadequate educational preparation for college.

(2) Many small companies don’t have sufficient resources to take on extra work because they can’t just run out and hire the extra people they need, perhaps they’re in a small town or perhaps they haven’t extra capital to invest before they are paid for the new work.

(3) Large companies are feeling pressure to outsource work even when they can see that the costs are larger than those who follow this path are willing to admit.

(4) There is a growing need in this country for well-trained technical people, while enrollments in standard engineering programs are decreasing because students feel that to choose these majors means that their jobs will just be candidates for outsourcing.

New Mexico Highlands University (NMHU) is a 4-year public university in Las Vegas, New Mexico. (Whatever joke you’re thinking about for "Las Vegas," trust me, they’ve already heard it!) The school is relatively new, founded in 1893. Students can get Bachelor's and Master's degrees in a variety of subjects. There are approximately 2,000 undergraduates and 800 graduate students. A little over half are female and most are Hispanic.

Its mission statement has some twists that might prepare you for an unusual solution or two:

NMHU is committed to excellence in teaching, discovering, preserving and applying knowledge and responding to new opportunities for teaching, learning, research, and public service created by a changing environment.

NMHU provides personal attention to students from distinctive cultural, socioeconomic, linguistic, geographic, religious, and educational backgrounds.

NMHU offers sensitive admissions programs and challenging academic programs that encourage all students to reach their full potential.

NMHU is committed to programs that focus on its multiethnic student body, especially the rich heritage of Hispanic and Native American cultures and clearly perceives that its success depends upon an appreciation of the region’s cultural and linguistic identities.

NMHU began its focus in the 1920s as the first institution of higher education serving the Hispanic population and remains true to that mission today. NMHU was chartered to service the people of northern New Mexico, which comprises a large rural population characterized by diverse cultural, economic, linguistic, and educational backgrounds, ranging from 82% Hispanic (Mora County) to 75% Native American (McKinley County). General poverty is a fact of life for this region and its students. The 2000 US Census ranked New Mexico as the poorest state in the nation, and the poverty rate in San Miguel County, home of NMHU, is 24%—double the national average. At 7.8%, unemployment is presently one fifth higher than the US average (NM Dept of Labor 2003). Given these demographics, the employment possibilities engendered by the program I visited offer hope and recognition for the hard work the students will have accomplished by the time they receive their degrees.

You can see more about this program on the NMHU web site: http://www.nmhu.edu/softwaredevelopment/

The average class size for undergraduates in the lower division is an astounding 21.20 students per class; for undergraduates in the upper division, 8.11 students per class, for graduate classes, 6.63 students per class. Think about those undergraduate classes you might have had with hundreds of students packed into an auditorium and a non-native-English-speaking graduate student lecturing in a whisper. What a different learning environment!

This place sounds like a wonderful setting for students who might not otherwise even dream of a college education. The mission of the institution and the small class size would tend to attract those faculty members who think about teaching rather than high-powered research. This is a school where people care about the students.

But what about the other problems in the list? How can this special school meet the needs of small companies, large companies, and what’s the tie-in with outsourcing?

Let’s move down from the institutional level to one unique department at this university and meet Dave West, former professor at the University of St. Thomas in Minneapolis-St. Paul, MN. Dave moved to NMHU to start a different kind of software development program. It is an apprenticeship program where students are paid for working on real projects for real customers. The emphasis is on competencies instead of courses. The approach is agile, in the best sense, in that it is people-centric. Everyone involved in this effort believes that developers and users are people first.

The Bachelor of Software Development Apprenticeship (SDA) program at NMHU is the realization of Dave’s long-held goal: teaching students in a community to use software development as a lever to change the world instead of using it as a tool for doing the same things people have always done more quickly and accurately. This reflects his belief that the key to improvement is better people rather than more rigidly specified processes.

The SDA program, housed in the School of Business Administration, started in August 2004, in a suite of three rooms optimized for agile development. Start-up difficulties related to equipment, physical facilities and staffing resulted in an attrition rate of almost 50% that first semester. The nine apprentices who continued into the second semester, however, have seen their numbers more than double and have increased their mastery of technology through group learning activities, interaction with customers, mentorship from object-oriented professionals, and instruction from nationally-known speakers.

They meet daily, often without faculty direction, to plan and construct software that has been ordered by one of the program’s three customers: New Mexico’s Office of the State Engineer Water Rights Management program, Tapetes de Lana (a mini-conglomerate with a comparable social focus), and the West Las Vegas School District. The needs of these three projects drive the content presented to the students and motivate the exercises assigned to them.

Now can you see the connection? The students do real work and are paid to learn. They don’t earn a lot of money, but for this part of the world and these individuals, it’s a strong incentive to stay with the program. It helps them with their finances and provides a clear path to a job.

Now, let’s look at small businesses in the area. Suppose you were running a small software company and had a chance to pick up some new work, but a couple of things were holding you back. (1) The new project would require more people than you have on staff now. You don’t want to run out and hire and train new people, since you’re not sure how things will pan out. If things don’t go well, you sure don’t need extra mouths to feed. (2) The new assignment involves Java. This is a new, cool programming language, but no one in your small company has any experience with it. Yes, you’d like to take the time to send some of your folks for training, but you can’t afford that just now.

What if you, the small business owner, could find a near-by school with an apprenticeship program where students are not experts, but they’ve been learning the latest, coolest technologies. You can employ them for the short-term and they are cheap. They’re enthusiastic and energetic and this near-by school will even let you come on-site a few days a week and work with the students. At the end of the project, your developers will have learned alongside their student colleagues—free training! To me, this looks like a win-win!

What about large companies? True, they might also have some of the training and staffing issues and could take advantage of the apprenticeship program, but large companies are facing the pressure to outsource. This is one of those "everyone else is doing it, so we might be missing out on something if we don’t get on board" items. Every CEO of every major corporation has an outsourcing proponent on his (or her) staff. The arguments never end.

Outsourcing is cheaper (look at costs for developers in the U.S. as compared with developers in India or China). We can not only cut costs, but we can also expand our market share. Think of the incredible number of widgets we can sell in China!

It’s hard to find an answer for this. Unless you’ve been looking carefully at the reports of the real costs of outsourcing—it does appear on the surface—that it would lead to significant savings. It’s hard to resist the siren call to line up behind the others.

What if you were a large organization in the vicinity of Albuquerque, New Mexico? Maybe a large government agency? What if a near-by school offered you the chance to "outsource" your work at essentially minimum wage but still keep that work within the U.S.? Such a deal! Not only that, but by increasing the wage-earning potential of communities in your state and thereby raising the standard of living for residents, you might also increase your market share right here at home!

Finally, let’s look at that last serious issue on the problem list. As outsourcing robs resident technical employees of their jobs, there is less incentive for college bound students to consider technical areas that are the targets of this kind of job loss. In the long run, who will suffer for this? If we outsource all our technical needs, we put our country at tremendous risk. What goes around comes around. Yes, there are discrepancies in the global market. I do believe that we truly deal in a global marketplace. Inequities will work themselves out. But we are suffering from inequities at home! Surely those must be addressed by some kind of balancing act!

Just as we’ve seen that the solutions proposed for other global problems are all technical, the solutions that are being proposed for the outsourcing problem are technical and legal. Let’s just make enough new rules to keep bad things from happening. We sure don’t want things to be "looser"—when you turn people loose, who knows what they will do!

I think Dave West is concerned about these inequities. I think that’s why Dave West is giving his time and energy to this program. I think that’s why Pam Rostal, one of Dave’s graduate students at St. Thomas, moved to Las Vegas, leaving her family in Minneapolis for a couple of years, to help jumpstart this program. I think that’s why I went to visit NMHU to see for myself what unusual solution has been implemented by a small handful of people who care.

The appealing essence of this solution is that it’s humane! It speaks to my common sense. Instead of more rules, restrictions and barriers, let’s do what we can to make the world a better place—for all of us!

Just as I was finishing this article, I read "Made in lower-cost America" in CNET news: http://news.com.com/Made+in+lower-cost+America/2100-1022_3-5562732.html

In this piece, I stumbled across the term "homeshoring" to describe the use by IT companies of workers in smaller cities or rural areas, where the cost of living and doing business is cheaper. In fact, one company notes that "fees are about the same as the overall cost of using an Indian outsourcer, if you consider factors such as communication costs, travel expenses, and inconvenience." Maybe this unusual solution isn’t so unusual after all!

I like that famous quote attributed to anthropologist Margaret Mead: "Never doubt that a small, committed group of people will change the world. Indeed, it's the only thing that ever has."

I had planned to end this series on unusual solutions with this article, but I’ve been hearing from readers and uncovering even more intriguing things to report, so perhaps I’ll do just one more column on this! Keep those cards and letters coming! Thanks for your interest and support!

About the Author http://www.lindarising.org risingl@acm.org Linda Rising has a Ph.D. from Arizona State University in the area of object-based design metrics. Her background includes university teaching as well as work in industry in telecommunications, avionics, and strategic weapons systems. She is the author of numerous articles and has published three books: Design Patterns in Communications, The Pattern Almanac 2000, and A Patterns Handbook. Follow this link for information regarding her latest book "Fearless Change: patterns for introducing new ideas". http://www.cs.unca.edu/~manns/intropatterns.html

[ Back to Top ]